1 2018-04-30 Jason Merrill <jason@redhat.com>
3 * tree.c (build_clobber): New.
5 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
7 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9 * diagnostic-show-locus.c (layout::layout): Update for
10 location_get_source_line returning a char_span.
11 (struct char_span): Move to input.h.
12 (struct correction): Update for fields in char_span becoming
14 (struct source_line): Update for location_get_source_line
15 returning a char_span.
16 (layout::print_line): Likewise.
17 * edit-context.c (edited_file::print_content): Likewise.
18 (edited_file::print_diff_hunk): Likewise.
19 (edited_file::print_run_of_changed_lines): Likewise.
20 (edited_file::get_num_lines): Likewise.
21 (edited_line::edited_line): Likewise.
22 * final.c (asm_show_source): Likewise.
23 * input.c (location_get_source_line): Convert return type
24 from const char * to char_span, losing the final "line_len"
26 (dump_location_info): Update for the above.
27 (get_substring_ranges_for_loc): Likewise. Use a char_span
28 when handling the literal within the line.
29 (test_reading_source_line): Update for location_get_source_line
30 returning a char_span.
31 * input.h (class char_span): Move here from
32 diagnostic-show-locus.c, converting from a struct to a class.
33 Make data members private.
34 (char_span::operator bool): New.
35 (char_span::length): New.
36 (char_span::get_buffer): New.
37 (char_span::operator[]): New.
38 (char_span::subspan): Make const.
39 (char_span::xstrdup): New.
40 (location_get_source_line): Convert return type from const char *
41 to char_span, losing the final "line_size" param.
43 2018-04-30 Jan Hubicka <jh@suse.cz>
45 * lto-wrapper.c (ltrans_priorities): New static var.
47 (run_gcc): Read priorities and if doing parallel build order
50 2018-04-30 David Malcolm <dmalcolm@redhat.com>
52 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
54 2018-04-30 Richard Biener <rguenther@suse.de>
56 * tree-cfg.c (verify_address): Remove base argument, add
57 flag whether to check TREE_ADDRESSABLE and do that.
58 (verify_expr): Remove.
59 (verify_types_in_gimple_reference): Add pieces from verify_expr.
60 (verify_gimple_assign_single): Likewise.
61 (verify_gimple_switch): Likewise.
62 (verify_expr_location_1): Dereference tp once. Add (disabled)
63 piece from verify_expr.
64 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
66 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
68 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
70 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
72 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
73 (small_data_pattern): Likewise.
74 (arc_rewrite_small_data): Likewise.
75 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
76 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
77 (get_symbol_alignment): New function.
78 (legitimate_small_data_address_p): Likewise.
79 (legitimate_scaled_address): Update, call
80 legitimate_small_data_address_p.
81 (output_sdata): New static variable.
82 (arc_print_operand): Update how we handle small data operands.
83 (arc_print_operand_address): Likewise.
84 (arc_legitimate_address_p): Update, use
85 legitimate_small_data_address_p.
86 (arc_rewrite_small_data_p): Remove.
87 (arc_rewrite_small_data_1): Likewise.
88 (arc_rewrite_small_data): Likewise.
89 (small_data_pattern): Likewise.
90 (compact_sda_memory_operand): Update to use
91 legitimate_small_data_address_p and get_symbol_alignment.
92 (prepare_move_operands): Don't rewite sdata pattern.
93 (prepare_extend_operands): Remove.
94 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
96 (zero_extendqisi2): Likewise.
97 (zero_extendhisi2): Likewise.
98 (extendqihi2): Likewise.
99 (extendqisi2): Likewise.
100 (extendhisi2): Likewise.
104 * config/arc/constraints.md (Usd): Change it to memory constraint.
106 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
108 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
109 as source of std instructions.
110 * config/arc/arc.md (movsi_insn): Update pattern predicate to
111 allow 6-bit constants as source for store instructions.
112 (movdi_insn): Update instruction pattern to allow 6-bit constants
113 as source for store instructions.
115 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
117 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
119 2018-04-30 Nathan Sidwell <nathan@acm.org>
120 Sandra Loosemore <sandra@codesourcery.com>
122 * dumpfile.c (dump_open): Allow '-' for stdout.
123 * doc/invoke.texi (Developer Options): Document dump filename
124 determination early. Document stdin/stdout selection.
126 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
128 Microblaze Target: PIC data text relative
130 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
131 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
133 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
134 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
135 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
136 New addressing mode for data-text relative position indepenedent code.
137 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
138 'ADDRESS_SYMBOLIC_TXT_REL'.
139 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
140 (microblaze_legitimate_pic_operand): Exclude function calls from
141 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
142 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
144 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
145 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
146 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
147 for 'address + offset'.
148 (microblaze_expand_prologue): Add new function prologue call for
150 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
151 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
152 table in case of TARGET_PIC_DATA_TEXT_REL.
153 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
154 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
155 Add new macros 'UNSPEC_TEXT',
156 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
157 + exclude function calls from 'UNSPEC_PLT' in case of data text
159 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
160 new target hook for generating address diff vector tables in case of
162 * doc/tm.texi : Regenerate.
163 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
164 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
165 of addr diff vector generation.
166 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
167 target hook definition.
168 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
169 Add default function for generate_pic_addr_diff_vec -> flag_pic.
170 * doc/invoke.texi (Add new pic option): Add new microblaze pic
171 option for data text relative.
173 2018-04-30 Richard Biener <rguenther@suse.de>
175 * tree-chrec.h (evolution_function_is_constant_p): Remove
177 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
179 2018-04-30 Richard Biener <rguenther@suse.de>
182 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
184 2018-04-30 Richard Biener <rguenther@suse.de>
186 PR tree-optimization/28364
187 PR tree-optimization/85275
188 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
189 copying first exit test.
191 2018-04-28 Mark Wielaard <mark@klomp.org>
193 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
195 (dwarf_AT): Handle DW_AT_addr_base.
196 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
198 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
201 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
202 (*ashl<dwi>3_doubleword_mask_1): Ditto.
203 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
204 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
206 2018-04-28 Richard Biener <rguenther@suse.de>
208 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
209 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
210 to reflect use. Only add interesting stmts.
212 2018-04-27 Martin Jambor <mjambor@suse.cz>
215 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
216 the jump function allows for passing through aggregate values.
218 2018-04-27 David Malcolm <dmalcolm@redhat.com>
220 * input.h (in_system_header_at): Convert from macro to inline
222 (from_macro_expansion_at): Likewise.
223 (from_macro_definition_at): Likewise.
225 2018-04-27 Jeff Law <law@redhat.com>
227 * config.gcc: Mark tile* targets as deprecated/obsolete.
229 2018-04-27 Richard Biener <rguenther@suse.de>
231 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
234 2018-04-27 Richard Biener <rguenther@suse.de>
236 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
238 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
240 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
241 with Yd constraint. Set "preferred_for_speed" attribute from
242 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
244 (*movdi_internal): Ditto.
245 (movti_interunit splitters): Remove
246 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
247 (movdi_interunit splitters): Ditto.
248 * config/i386/constraints.md (Ye): Remove.
249 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
251 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
254 * config/aarch64/constraints.md (Usg): Limit to 31.
257 2018-04-27 Jakub Jelinek <jakub@redhat.com>
259 PR tree-optimization/85529
260 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
261 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
262 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
263 zero extension or masking of the MSB bit.
264 (optimize_range_tests): Add FIRST_BB argument, pass it through
265 to optimize_range_tests_var_bound.
266 (maybe_optimize_range_tests, reassociate_bb): Adjust
267 optimize_range_tests callers.
269 2018-04-26 Richard Biener <rguenther@suse.de>
270 Jakub Jelinek <jakub@redhat.com>
272 * cgraph.h (symbol_table): Just declare debug method here.
273 * symtab.c (symbol_table::debug): Define.
275 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
277 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
279 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
281 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
282 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
283 (*movdi_internal): Substitute Yi and Yj constraint with x
284 and Ym and Yn constraint with y constraint. Update "isa"
285 attribute and set "preferred_for_speed" attribute from
286 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
287 (*movsi_internal): Ditto.
288 (*movdf_internal): Ditto.
289 (*movsf_internal): Ditto.
290 (*zero_extendsidi2): Ditto.
291 * config/i386/sse.md (vec_set<mode>_0): Ditto.
292 (sse2_loadld): Ditto.
293 (*vec_extract<ssevecmodelower>_0): Ditto.
294 (*vec_extractv4si_0_zext_sse4): Ditto.
295 (vec_concatv2di): Ditto.
296 (*vec_dup<mode>): Ditto.
297 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
298 * config/i386/constraints.md (Yi): Remove.
303 2018-04-26 Nathan Sidwell <nathan@acm.org>
305 * dumpfile.c (dump_open): New.
306 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
307 (dump_finish): Detect stdio/stderr by value not name.
309 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
311 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
313 2018-04-26 Tom de Vries <tom@codesourcery.com>
316 * config/nvptx/nvptx.c (verify_neutering_jumps)
317 (verify_neutering_labels): New function
318 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
320 2018-04-26 Tom de Vries <tom@codesourcery.com>
323 * config/nvptx/nvptx.c (needs_neutering_p): New function.
324 (nvptx_single): Use needs_neutering_p to skip over insns that do not
327 2018-04-26 Richard Biener <rguenther@suse.de>
328 Tom de Vries <tom@codesourcery.com>
331 * lto-streamer-out.c (output_function): Fixup loops if required to match
332 discovery done in the reader.
334 2018-04-26 Richard Biener <rguenther@suse.de>
336 PR tree-optimization/85116
337 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
338 have a loop exit from the single latch predecessor. Remove
339 case of header with just condition.
340 (ch_base::copy_headers): Exclude infinite loops from any
342 (pass_ch::execute): Record exits.
344 2018-04-26 Richard Biener <rguenther@suse.de>
346 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
347 prologue cost vector and pass it to vect_get_load_cost.
348 (vect_get_peeling_costs_all_drs): Likewise.
349 (vect_peeling_hash_get_lowest_cost): Likewise.
350 (vect_enhance_data_refs_alignment): Likewise.
352 2018-04-26 Richard Biener <rguenther@suse.de>
355 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
356 checking of integer<->pointer conversions.
357 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
358 sign-/zero-extending pointer types.
359 (expand_omp_for_static_chunk): Likewise.
361 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
362 Jean Lee <xiaoyur347@gmail.com>
364 * config/mips/mips.c (mips_asan_shadow_offset): New function.
365 (TARGET_ASAN_SHADOW_OFFSET): Define.
366 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
367 true for -fsanitize=address.
369 2018-04-25 Mark Wielaard <mark@klomp.org>
371 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
374 2018-04-25 Jakub Jelinek <jakub@redhat.com>
376 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
377 than "alu", remove explicit "memory" and "imm_disp" attributes.
378 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
381 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
382 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
385 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
388 * config/i386/i386.c (ix86_expand_builtin): Change memory
389 operand to XI, extend p0 to Pmode.
390 * config/i386/i386.md: Change unspec volatile and operand
391 1 mode to XI, change operand 0 mode to P.
393 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
395 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
396 GET_MODE_MASK before any checking.
397 (nds32_can_use_bset_p): Likewise.
398 (nds32_can_use_btgl_p): Likewise.
400 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
402 * config/nds32/nds32-doubleword.md: New define_split pattern for
403 illegal register number.
405 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
407 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
409 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
411 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
413 2018-04-25 Richard Biener <rguenther@suse.de>
415 * lto-streamer.h (LTO_major_version): Bump to 8.
417 2018-04-25 Jakub Jelinek <jakub@redhat.com>
419 * BASE-VER: Set to 9.0.0.
421 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
423 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
424 in __abskf2 and __powikf2.
426 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
429 * config/aarch64/constraints.md (Usg, Usj): New constraints.
430 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
431 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
432 Use the above on operand 2. Reindent.
433 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
435 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
438 * common/config/i386/i386-common.c (ix86_handle_option): Don't
440 * config/i386/i386.opt (mcet): Removed.
441 * doc/install.texi: Remove -mcet documentation.
442 * doc/invoke.texi: Likewise.
444 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
447 * doc/install.texi: Remove -mcet from bootstrap-cet.
449 2018-04-24 Jakub Jelinek <jakub@redhat.com>
452 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
453 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
457 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
458 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
459 containing a CONST_VECTOR.
461 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
463 * doc/install.texi: Update newlib dependency for nvptx.
465 2018-04-24 Jakub Jelinek <jakub@redhat.com>
468 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
469 instead of INTVAL when shifting x left.
471 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
473 PR tree-optimization/85478
474 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
475 vect_grouped_store_supported for single element vectors.
477 2018-04-24 Richard Biener <rguenther@suse.de>
480 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
481 load cost increase to the case of non-constant step.
483 2018-04-24 Jakub Jelinek <jakub@redhat.com>
486 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
487 destination if any_malformed_asm.
489 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
492 * expr.c (store_field): In the bitfield case, if the value comes from
493 a function call and is returned in registers by means of a PARALLEL,
494 do not change the mode of the temporary unless BLKmode and VOIDmode.
496 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
498 PR rtl-optimization/85423
499 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
500 dependencies to debug insns when the previous insn is non-debug.
502 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
504 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
505 enums into a single definition.
506 (fls): Fix predicates and printing.
509 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
511 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
512 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
513 and short u6 immediate.
514 (check_if_valid_sleep_operand): Remove.
515 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
517 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
519 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
520 flag_always_save_lp condition.
521 * config/nds32/nds32.opt (malways-save-lp): New option.
523 2018-04-22 Shiva Chen <shiva0217@gmail.com>
525 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
526 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
527 * config/nds32/nds32.h
528 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
529 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
531 2018-04-22 Shiva Chen <shiva0217@gmail.com>
533 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
534 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
536 2018-04-22 Shiva Chen <shiva0217@gmail.com>
537 Chung-Ju Wu <jasonwucj@gmail.com>
539 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
541 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
542 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
544 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
546 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
548 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
550 * config/nds32/nds32-protos.h (nds32_data_alignment,
551 nds32_local_alignment): Declare.
552 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
553 nds32_local_alignment): New functions.
554 (TARGET_CONSTANT_ALIGNMENT): Define.
555 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
557 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
559 * config/nds32/nds32.c
560 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
561 (TARGET_MODES_TIEABLE_P): Likewise.
563 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
565 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
568 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
569 Chung-Ju Wu <jasonwucj@gmail.com>
571 * config/nds32/constants.md (unspec_volatile_element): Add enum values
572 for unaligned access.
573 * config/nds32/nds32-intrinsic.c: Implementation of expanding
575 * config/nds32/nds32-intrinsic.md: Likewise.
576 * config/nds32/nds32_intrinsic.h: Likewise.
577 * config/nds32/nds32.h (nds32_builtins): Likewise.
578 * config/nds32/nds32.opt (munaligned-access): New option.
579 * config/nds32/nds32.c (nds32_asm_file_start): Display
580 flag_unaligned_access status.
582 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
584 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
585 -mno-relax is present.
586 * config/riscv/linux.h (LINK_SPEC): Ditto.
588 2018-04-20 Martin Sebor <msebor@redhat.com>
591 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
593 (gimple_fold_builtin_stxcpy_chk): Same.
594 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
596 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
599 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
600 __powikf2 when long double is IEEE 128-bit.
602 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
604 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
605 step to make sure stack always aligned.
607 2018-04-20 Carl Love <cel@us.ibm.com>
610 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
613 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
614 Tom de Vries <tom@codesourcery.com>
617 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
618 Emit insns for calls too.
619 (nvptx_find_par): Always look for worker-level predecessor insn.
620 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
622 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
623 (nvptx_process_pars): Propagate frames for calls.
625 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
628 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
630 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
631 (ix86_handle_option): Don't handle OPT_mibt.
632 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
634 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
636 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
637 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
638 (ix86_target_macros): Define __CET__ with flag_cf_protection
640 * config/i386/i386.c (isa2_opts): Remove -mibt.
641 * config/i386/i386.h (TARGET_IBT): Removed.
642 (TARGET_IBT_P): Likewise.
643 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
644 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
645 * config/i386/i386.opt (mcet): Update help message.
648 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
649 -mcet as an alias for -mshstk.
651 2018-04-20 Richard Biener <rguenther@suse.de>
654 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
655 complexity by forcing a single use of the multiply operand.
657 2018-04-20 Martin Jambor <mjambor@suse.cz>
660 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
661 recursion dependency to only apply to non-clones.
663 2018-04-20 Martin Jambor <mjambor@suse.cz>
666 * ipa-cp.c (create_specialized_node): Check that clones of
667 self-recursive edges exist during IPA-CP.
669 2018-04-19 Toon Moene <toon@moene.org>
671 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
674 2018-04-19 Jakub Jelinek <jakub@redhat.com>
676 PR tree-optimization/85467
677 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
678 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
679 VECTOR_CST element to type.
681 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
684 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
685 * config/i386/i386.md (builtin_setjmp_setup): Removed.
686 (builtin_longjmp): Likewise.
687 (save_stack_nonlocal): New pattern.
688 (restore_stack_nonlocal): Likewise.
690 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
693 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
694 Replace ASM_OUTPUT_LABEL with fprintf.
696 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
699 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
700 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
701 * config/i386/i386-c.c (ix86_target_macros_internal): Also
702 define __IBT__ and __SHSTK__ for -fcf-protection.
703 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
705 (ix86_trampoline_init): Likewise.
706 (x86_output_mi_thunk): Likewise.
707 (ix86_notrack_prefixed_insn_p): Likewise.
708 (ix86_option_override_internal): Don't disallow -fcf-protection.
709 * config/i386/i386.md (rdssp<mode>): Also enable for
711 (incssp<mode>): Likewise.
712 (nop_endbr): Likewise.
713 * config/i386/i386.opt (mcet): Change help message to built-in
717 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
718 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
719 enable CET built-in functions.
721 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
723 * common/config/i386/i386-common.c
724 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
725 OPTION_MASK_ISA_MOVDIRI_UNSET,
726 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
727 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
728 * config.gcc (movdirintrin.h): New header.
729 * config/i386/cpuid.h (bit_MOVDIRI,
730 bit_MOVDIR64B): New bits.
731 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
733 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
734 (VOID, PVOID, PCVOID)): New function types.
735 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
736 __builtin_ia32_directstoreu_u64,
737 __builtin_ia32_movdir64b): New builtins.
738 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
739 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
741 (ix86_valid_target_attribute_inner_p): Ditto.
742 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
743 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
744 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
745 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
746 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
747 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
748 (movdiri<mode>, movdir64b_<mode>): New.
749 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
750 * config/i386/immintrin.h: Include movdirintrin.h.
751 * config/i386/movdirintrin.h: New file.
752 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
754 2018-04-19 Richard Biener <rguenther@suse.de>
757 * cfg.c (clear_bb_flags): When loop state says we have
758 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
760 2018-04-19 Richard Biener <rguenther@suse.de>
762 PR tree-optimization/84737
763 * tree-vect-data-refs.c (vect_copy_ref_info): New function
764 copying restrict info.
765 (vect_setup_realignment): Use it.
766 * tree-vectorizer.h (vect_copy_ref_info): Declare.
767 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
768 the first DR to all generated stores.
769 (vectorizable_load): Likewise for loads.
771 2018-04-19 Jakub Jelinek <jakub@redhat.com>
773 PR tree-optimization/85446
774 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
775 the integral and pointer types to have the same precision.
777 * doc/install.texi: Document --disable-cet being the default and
780 2018-04-18 Martin Liska <mliska@suse.cz>
782 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
785 2018-04-18 Martin Liska <mliska@suse.cz>
788 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
791 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
792 arguments if they are comparable.
794 2018-04-18 Martin Liska <mliska@suse.cz>
797 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
800 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
803 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
806 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
807 ENDBR after calling __morestack.
809 2018-04-18 David Malcolm <dmalcolm@redhat.com>
812 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
813 by using gcc_base_ver to generate a gcc_driver_version, and use
814 it when generating GCC_DRIVER_NAME.
815 * configure: Regenerate.
817 2018-04-18 Jakub Jelinek <jakub@redhat.com>
820 * config.gcc: Obsolete powerpc*-*-*spe*.
822 2018-04-17 Jakub Jelinek <jakub@redhat.com>
825 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
826 (stabstr_D): Change type of unum from unsigned int to
827 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
830 2018-04-17 Jim Wilson <jimw@sifive.com>
833 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
834 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
835 Set arg_pointer_offset after using pretend_args_size.
837 2018-04-17 Jakub Jelinek <jakub@redhat.com>
839 PR rtl-optimization/85431
840 * dse.c (record_store): Ignore zero width stores.
843 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
844 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
845 __builtin_stack_restore rather than after it.
846 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
847 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
848 argument instead of virtual_dynamic_stack_rtx.
850 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
852 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
854 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
855 Add note to error message to explain internal mapping of overloaded
856 built-in function name to non-overloaded built-in function name.
857 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
860 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
863 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
864 where the inputs overlap with the output.
866 2018-04-17 Jakub Jelinek <jakub@redhat.com>
868 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
869 (=v, v) alternative and explicit "memory" attribute.
870 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
871 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
873 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
874 "sselog1" type instead of "sselog".
875 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
876 "sselog". Remove explicit "memory" attribute.
877 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
878 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
880 (vec_extract_hi_v32hi): Merge all alternatives into one, use
881 "sselog1" type instead of "sselog". Remove explicit "memory"
883 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
884 use "sselog1" type instead of "sselog". Remove explicit "memory"
886 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
887 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
889 (vec_extract_hi_v64qi): Merge all alternatives into one, use
890 "sselog1" type instead of "sselog". Remove explicit "memory"
892 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
893 use "sselog1" type instead of "sselog". Remove explicit "memory"
897 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
900 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
903 2018-04-17 Martin Jambor <mjambor@suse.cz>
906 * ipa-cp.c (create_specialized_node): Call
907 expand_all_artificial_thunks if necessary.
909 2018-04-17 Martin Liska <mliska@suse.cz>
912 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
913 in message, remote space in between '_G' and '('.
915 2018-04-17 Jakub Jelinek <jakub@redhat.com>
918 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
919 avx512f_vmcmp<mode>3<round_saeonly_name>,
920 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
921 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
922 avx512f_rndscale<mode><round_saeonly_name>,
923 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
924 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
925 Use %<iptr>2 instead of %2 for -masm=intel.
926 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
927 avx512f_vcvttss2usi<round_saeonly_name>,
928 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
930 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
931 avx512f_vcvttsd2usi<round_saeonly_name>,
932 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
933 Use %q1 instead of %1 for -masm=intel.
934 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
935 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
936 of %3 for -masm=intel.
937 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
939 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
941 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
943 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
944 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
946 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
948 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
949 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
950 %g1 and one with %0 and %1.
951 (avx512er_vmrcp28<mode><round_saeonly_name>,
952 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
954 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
955 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
956 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
957 of %0 and %{%4%} for -masm=intel.
958 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
959 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
960 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
961 order of %0 and %{%5%}%{z%} for -masm=intel.
963 2018-04-17 Jan Hubicka <jh@suse.cz>
966 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
968 2018-04-17 Martin Liska <mliska@suse.cz>
971 * multiple_target.c (create_dispatcher_calls): Set apostrophes
972 for target_clone error message. Make default implementation
973 clone to be a local declaration.
974 (separate_attrs): Add new argument and check for an empty
976 (expand_target_clones): Handle it.
977 (ipa_target_clone): Make redirection just for target_clones
980 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
981 Tom de Vries <tom@codesourcery.com>
984 * omp-expand.c (expand_oacc_for): Add dummy false branch for
985 tiled basic blocks without omp continue statements.
987 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
990 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
991 vec_extract expression as having side effects to make sure it gets
994 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
997 * config/i386/i386.c (get_builtin_code_for_version): Check
1000 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1003 * gcc/config.gcc: Support "skylake".
1004 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1006 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1007 (processor_target_table): Add "skylake".
1008 (ix86_option_override_internal): Add "skylake".
1009 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1010 PROCESSOR_CANNONLAKE.
1011 (get_builtin_code_for_version): Fix priority for
1012 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1013 PROCESSOR_SKYLAKE-AVX512.
1014 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1015 (processor_type): Add PROCESSOR_SKYLAKE.
1017 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1018 Jason Merrill <jason@redhat.com>
1021 * convert.c (convert_to_integer_1): Use direct recursion for
1022 enumeral types and types with a precision less than the number
1023 of bits in their mode.
1025 2018-04-16 Julia Koval <julia.koval@intel.com>
1028 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1029 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1031 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1034 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1035 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1036 and -mno-direct-move.
1038 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1041 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1042 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1043 Ensure negative shifts result in {0}.
1045 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1047 PR rtl-optimization/79916
1048 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1049 regs (if any) to define how to gnerate SD moves when LRA is in
1052 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1054 PR rtl-optimization/85393
1055 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1056 * except.c (expand_dw2_landing_pad_for_region): Make static.
1057 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1058 a label and unconditional jump to old_bb, rather than
1059 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1062 PR rtl-optimization/85376
1063 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1064 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1065 instead of a specific value.
1067 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1068 Bin Cheng <bin.cheng@arm.com>
1070 PR tree-optimization/82965
1071 PR tree-optimization/83991
1072 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1073 by_profile_only parameter.
1074 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1075 information if the loop was predicted to iterate too many times.
1076 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1078 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1081 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1084 2018-04-13 Martin Liska <mliska@suse.cz>
1085 Jakub Jelinek <jakub@redhat.com>
1088 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1089 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1090 * builtins.c (expand_builtin_memory_copy_args): Use
1091 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1092 handle dest_addr == pc_rtx.
1094 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1097 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1098 asked to not generate direct moves.
1099 (fix_trunc<mode>si2_stfiwx): Similar.
1100 (fix_trunc<mode>si2_internal): Similar.
1102 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1105 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1106 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1107 lookup if dest in some wider mode is known to be const0_rtx and
1108 if so, record permanent equivalence for it to be ZERO_EXTEND of
1109 the narrower mode destination.
1111 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1113 * lto-streamer-out.c (output_function): Revert 259346.
1114 * omp-expand.c (expand_oacc_for): Likewise.
1116 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1118 PR rtl-optimization/85354
1119 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1120 * sel-sched.c (sel_global_init): ... here.
1122 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1125 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1126 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1127 mode for PE-COFF targets.
1128 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1129 (i386_pe_asm_lto_end): Likewise.
1130 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1131 (TARGET_ASM_LTO_END): Likewise.
1132 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1133 (i386_pe_asm_lto_start): New function.
1134 (i386_pe_asm_lto_end): Likewise.
1136 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1137 Richard Biener <rguenther@suse.de>
1140 * lto-streamer-out.c (output_function): Fix CFG loop state before
1142 * omp-expand.c (expand_oacc_for): Handle calls to internal
1143 functions like regular functions.
1145 2018-04-12 Richard Biener <rguenther@suse.de>
1148 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1149 for the early LTO debug to properly generate references to it
1150 during DIE emission. Do not re-use that for the skeleton for
1152 (dwarf2out_early_finish): Likewise.
1154 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1157 * config/i386/sse.md
1158 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1159 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1160 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1161 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1162 and output is a reg, avoid creating invalid lowpart subreg, but
1163 instead split into a 512-bit move. Don't split if not AVX512VL,
1164 input is xmm16+ reg and output is a mem.
1165 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1166 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1167 xmm16+ reg and output is a mem.
1169 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1171 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1172 also for flag_dwarf2_cfi_asm.
1174 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1176 PR rtl-optimization/85342
1177 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1178 a bool scalar var inside of the loop instead. Don't try to update
1179 recog_data.operand after failed apply_change_group.
1181 2018-04-12 Tom de Vries <tom@codesourcery.com>
1184 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1185 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1186 array with flexible array member as array without given dimension.
1187 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1188 argument for undefined param to true.
1190 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1193 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1194 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1195 from PowerPC section.
1196 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1197 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1198 help text that is too long.
1199 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1200 help text that is too long.
1201 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1202 help text that is too long.
1204 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1206 * config/alpha/alpha.md (stack_probe_internal): Rename
1207 from "probe_stack". Update all callers.
1209 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1211 PR rtl-optimization/84566
1212 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1213 sched_macro_fuse_insns.
1215 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1218 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1219 (compute_block_dependences): ... from here.
1221 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1223 PR tree-optimization/85331
1224 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1225 from int to HOST_WIDE_INT.
1227 2018-04-11 Martin Jambor <mjambor@suse.cz>
1230 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1231 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1232 not the same as the source val.
1233 (cgraph_edge_brings_value_p): New parameter.
1234 (gather_edges_for_value): Pass destination value to
1235 cgraph_edge_brings_value_p.
1236 (perhaps_add_new_callers): Likewise.
1237 (get_info_about_necessary_edges): Likewise and exclude values brought
1238 only by self-recursive edges.
1239 (create_specialized_node): Redirect only clones of self-calling edges.
1240 (+self_recursive_pass_through_p): New function.
1241 (find_more_scalar_values_for_callers_subset): Use it.
1242 (find_aggregate_values_for_callers_subset): Likewise.
1243 (known_aggs_to_agg_replacement_list): Removed.
1244 (decide_whether_version_node): Re-calculate known constants for all
1245 remaining context clones.
1247 2018-04-11 Richard Biener <rguenther@suse.de>
1250 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1251 from early DWARF output.
1252 (dwarf2out_early_finish): Output line info unconditionally into
1253 early DWARF and add reference to it.
1255 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1258 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1259 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1260 other than V2DFmode using iptr mode attribute.
1261 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1263 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1265 PR rtl-optimization/84659
1266 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1268 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1271 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1273 (output_loc_list): Pass address of a dummy size variable even in the
1274 locview handling loop.
1275 (index_location_lists): Add comment on why skip_loc_list_entry can't
1278 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1281 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1284 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1287 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1288 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1289 and -mstring-compare-inline-limit.
1291 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1294 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1295 for stack clash protection in a register whenever we need it to be in
1298 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1300 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1301 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1303 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1306 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1308 (mlong-double-): Ditto.
1309 * config/rs6000/sysv4.opt (msdata=): Ditto.
1310 (mtls-size=): Ditto.
1312 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1314 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1315 erroneous entries for
1316 "vector int vec_ldl (int, long int *)", and
1317 "vector unsigned int vec_ldl (int, unsigned long int *)".
1318 Add comments and entries for
1319 "vector bool char vec_ldl (int, bool char *)",
1320 "vector bool short vec_ldl (int, bool short *)",
1321 "vector bool int vec_ldl (int, bool int *)",
1322 "vector bool long long vec_ldl (int, bool long long *)",
1323 "vector pixel vec_ldl (int, pixel *)",
1324 "vector long long vec_ldl (int, long long *)",
1325 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1326 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1327 type tree bool_long_long_type_node and correct definition of
1328 bool_V2DI_type_node to make reference to this new type tree.
1329 (rs6000_mangle_type): Replace erroneous reference to
1330 bool_long_type_node with bool_long_long_type_node.
1331 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1332 comments to emphasize sign distinctions for char and int types and
1333 replace RS6000_BTI_bool_long constant with
1334 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1335 use of RS6000_BTI_pixel.
1336 (bool_long_type_node): Remove this macro definition.
1337 (bool_long_long_type_node): New macro definition
1339 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1341 PR rtl-optimization/85300
1342 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1343 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1344 simplify_unary_operation fails.
1346 2018-04-10 Martin Liska <mliska@suse.cz>
1348 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1349 cgraph_edge and ipa_ref.
1351 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1355 * config/i386/sse.md
1356 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1357 computation of the VEC_MERGE selector from mask.
1358 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1359 Fix decoding of the VEC_MERGE selector into mask.
1361 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1363 PR tree-optimization/85286
1364 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1366 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1368 * final.c (final_1): Set insn_last_address as well as
1369 insn_current_address.
1371 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1374 * explow.c (emit_stack_probe): Call validize_mem on memory location
1375 before passing it to gen_probe_stack. Create address operand and
1376 legitimize it for the probe_stack_address case.
1378 2018-04-09 Jan Hubicka <jh@suse.cz>
1381 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1382 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1383 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1385 (free_lang_data): Rebuild type inheritance graph.
1387 2018-04-09 Martin Sebor <msebor@redhat.com>
1389 * invoke.texi (-finline-small-functions): Mention other optimization
1391 (-findirect-inlining, -fpartial-inlining): Same.
1392 (-finline-functions-called-once): Same.
1393 (-freorder-blocks-and-partition): Same.
1395 2018-04-09 Jan Hubicka <jh@suse.cz>
1398 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1399 jumps; choose last target that matches the criteria (i.e.
1400 no partition changes for non-crossing jumps).
1401 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1402 support for redirecting crossing jumps to non-crossing.
1404 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1406 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1407 also for naked functions.
1409 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1411 * config/arc/arc.md (add_shift): New pattern.
1412 (add_shift2): Likewise.
1413 (sub_shift): Likewise.
1414 (sub_shift_cmp0_noout): Likewise.
1415 (compare_si_ashiftsi): Likewise.
1416 (xbfu_cmp0_noout): New combine pattern.
1417 (xbfu_cmp0"): Likewise.
1418 (movsi_set_cc_insn): Place the predicable variant first.
1419 (commutative_binary_cmp0_noout): Remove clobber.
1420 (commutative_binary_cmp0): New pattern.
1421 (noncommutative_binary_cmp0): Likewise.
1422 (noncommutative_binary_cmp0_noout): Likewise.
1423 (noncommutative_binary_comparison_result_used): Removed.
1424 (rsub_cmp0): New pattern.
1425 (rsub_cmp0_noout): Likewise.
1426 (extzvsi): Changed, keep only meaningful variants.
1427 (SQH, SEZ): New iterators.
1428 (SQH_postfix): New mode attribute.
1429 (SEZ_prefix): New code attribute.
1430 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1431 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1432 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1434 (noncommutative_operator): Check the availability of barrel
1437 2018-04-09 Richard Biener <rguenther@suse.de>
1439 PR tree-optimization/85284
1440 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1441 Only use the niter constraining form of simple_iv when the exit
1444 2018-04-09 Tom de Vries <tom@codesourcery.com>
1447 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1448 (define_expand "*memory_barrier"): New define_expand.
1449 (define_insn "memory_barrier"): New insn.
1451 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1453 PR rtl-optimization/80463
1454 PR rtl-optimization/83972
1455 PR rtl-optimization/83480
1457 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1458 correct producer for the insn.
1459 (tidy_control_flow): Fixup seqnos in case of debug insns.
1461 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1463 PR rtl-optimization/83913
1465 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1466 different sched-times when merging exprs.
1468 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1470 PR rtl-optimization/83962
1472 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1473 tidy_fallthru_edge and tidy_control_flow.
1475 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1477 PR rtl-optimization/83530
1479 * sel-sched.c (force_next_insn): New global variable.
1480 (remove_insn_for_debug): When force_next_insn is true, also leave only
1481 next insn in the ready list.
1482 (sel_sched_region): When the region wasn't scheduled, make another pass
1483 over it with force_next_insn set to 1.
1485 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1487 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1489 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1490 for interrupt control.
1491 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1492 functions for interrupt control.
1493 * config/nds32/nds32-intrinsic.md: Likewise.
1494 * config/nds32/nds32_intrinsic.h: Likewise.
1495 * config/nds32/nds32.h (nds32_builtins): Likewise.
1497 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1499 * config/nds32/nds32.c (nds32_init_machine_status,
1500 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1501 strict_aligned_p field.
1502 (nds32_expand_to_rtl_hook): New function.
1503 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1504 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1506 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1507 Chung-Ju Wu <jasonwucj@gmail.com>
1509 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1510 * config/nds32/nds32-n7.md: New file.
1511 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1512 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1514 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1515 * config/nds32/nds32.md (pipeline_model): Add n7.
1516 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1517 * config/nds32/pipelines.md: Include n7 settings.
1519 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1520 Chung-Ju Wu <jasonwucj@gmail.com>
1522 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1523 * config/nds32/nds32-e8.md: New file.
1524 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1525 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1527 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1528 * config/nds32/nds32.md (pipeline_model): Add e8.
1529 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1530 * config/nds32/pipelines.md: Include e8 settings.
1532 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1533 Chung-Ju Wu <jasonwucj@gmail.com>
1535 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1536 * config/nds32/nds32-n8.md: New file.
1537 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1538 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1540 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1541 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1542 * config/nds32/nds32.md (pipeline_model): Add n8.
1543 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1544 * config/nds32/pipelines.md: Include n8 settings.
1546 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1547 Chung-Ju Wu <jasonwucj@gmail.com>
1549 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1550 * config/nds32/nds32-n9-2r1w.md: New file.
1551 * config/nds32/nds32-n9-3r2w.md: New file.
1552 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1553 nds32_register_ports): New or modify for cpu n9.
1554 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1556 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1557 * config/nds32/nds32-utils.c: New file.
1558 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1559 TARGET_MUL_SLOW): Define.
1560 * config/nds32/nds32.md (pipeline_model): New attribute.
1561 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1562 New options that support cpu n9.
1563 * config/nds32/pipelines.md: Include n9 settings.
1564 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1566 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1568 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1569 information if necessary.
1570 (output_cond_branch_compare_zero): Likewise.
1571 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1572 (nds32_target_alignment): Refine for alignment.
1573 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1574 (FUNCTION_BOUNDARY): Modify.
1575 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1577 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1579 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1581 * config/nds32/constants.md (unspec_volatile_element): Add values for
1582 TLB operation and data prefetch.
1583 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1584 functions for TLB operation and data prefetch.
1585 * config/nds32/nds32-intrinsic.md: Likewise.
1586 * config/nds32/nds32_intrinsic.h: Likewise.
1587 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1588 (nds32_print_operand): Likewise.
1589 * config/nds32/nds32.h (nds32_builtins): Likewise.
1591 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1592 Andrew Pinski <pinsika@gcc.gnu.org>
1595 * match.pd: Use constant_boolean_node of correct type instead of
1596 boolean_true_node or boolean_false_node for simplifying
1597 pointer comparisons to zero.
1599 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1601 PR tree-optimization/80021
1602 * tree.c (verify_type_variant): Make error call in verify_variant_match
1603 translatable and remove final full stop.
1605 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1607 * config/nds32/constants.md (unspec_volatile_element): Add
1608 UNSPEC_VOLATILE_EH_RETURN.
1609 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1610 nds32_output_stack_pop): Support dwarf exception handling process.
1611 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1612 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1613 exception handling process.
1614 (nds32_compute_stack_frame): Likewise.
1615 (nds32_return_addr_rtx): Likewise.
1616 (nds32_initial_elimination_offset): Likewise.
1617 (nds32_expand_prologue): Likewise.
1618 (nds32_expand_epilogue): Likewise.
1619 (nds32_dynamic_chain_address): New function.
1620 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1622 (DYNAMIC_CHAIN_ADDRESS): Define.
1623 (EH_RETURN_DATA_REGNO): Define.
1624 (EH_RETURN_STACKADJ_RTX): Define.
1625 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1626 patterns for dwarf exception handling.
1628 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1630 * config/nds32/nds32.h: Clean up obsolete macros.
1632 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1634 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1635 Add enum values for particular instructions.
1636 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1637 particular intrinsic functions.
1638 * config/nds32/nds32-intrinsic.md: Likewise.
1639 * config/nds32/nds32_intrinsic.h: Likewise.
1640 * config/nds32/nds32.h (nds32_builtins): Likewise.
1641 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1642 (btst, ave): New patterns for particular instructions.
1644 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1646 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1647 Add enum values for atomic load/store and memory sync.
1648 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1650 * config/nds32/nds32-intrinsic.md: Likewise.
1651 * config/nds32/nds32_intrinsic.h: Likewise.
1652 * config/nds32/nds32.h (nds32_builtins): Likewise.
1654 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1656 PR tree-optimization/85257
1657 * fold-const.c (native_encode_vector): If not all elts could fit
1658 and off is -1, return 0 rather than offset.
1659 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1660 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1661 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1662 adjust buffer in native_interpret_expr call.
1664 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1666 * config/nds32/constants.md (unspec_volatile_element): Add cache
1667 control enum values.
1668 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1669 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1670 * config/nds32/nds32.c (nds32_cctl_names): New.
1671 (nds32_print_operand): Handle cache control register names.
1672 * config/nds32/nds32.h (nds32_builtins): New enum values.
1673 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1675 * config/nds32/nds32.md (type): Add mmu.
1676 * config/nds32/pipelines.md (simple_insn): Add mmu.
1678 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1680 * config/nds32/nds32.md (type): Remove call.
1681 * config/nds32/pipelines.md (simple_insn): Likewise.
1683 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1685 * config/nds32/constants.md (unspec_volatile_element): Add
1686 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1687 UNSPEC_VOLATILE_FMFCFG.
1688 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1689 description for fmfcfg and fmfcsr.
1690 (bdesc_1arg): Add fmtcsr.
1691 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1692 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1693 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1694 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1695 unspec_fmfcfg): New patterns.
1696 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1697 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1698 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1699 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1700 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1701 __nds32__fmfcfg): Define.
1703 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1705 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1706 intrinsic register names.
1707 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1708 intrinsic register enum values and macros.
1710 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1712 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1713 for load/store addressing form.
1714 (nds32_print_operand_address): Likewise.
1716 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1719 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1720 based on LABEL_REF. Remove useless assertion.
1721 (pic_address_needs_scratch): Fix formatting.
1722 (sparc_legitimize_pic_address): Minor tweaks.
1723 (sparc_delegitimize_address): Adjust assertion accordingly.
1724 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1725 into symbolic_operand.
1726 (movsi_high_pic_label_ref): Likewise.
1727 (movsi_lo_sum_pic_label_ref): Likewise.
1728 (movdi_pic_label_ref): Likewise.
1729 (movdi_high_pic_label_ref): Likewise.
1730 (movdi_lo_sum_pic_label_ref): Likewise.
1732 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1734 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1735 custom LIB_SPEC setup.
1737 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1738 Kito Cheng <kito.cheng@gmail.com>
1740 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1741 * config/riscv/freebsd.h: New.
1743 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1745 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1746 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1749 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1750 Kito Cheng <kito.cheng@gmail.com>
1752 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1753 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1754 * config/nds32/nds32-protos.h (nds32_output_call,
1755 nds32_output_return): Declare.
1756 * config/nds32/nds32.md: Refine all the call and return patterns.
1758 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1761 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1762 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1764 PR rtl-optimization/84872
1765 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1766 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1769 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1771 * expr.c (copy_blkmode_to_reg): Revert 254862.
1772 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1774 2018-04-06 Richard Biener <rguenther@suse.de>
1777 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1778 after seeing a component reference with an adjacent field. Treat
1779 refs to arrays at struct end of external decls similar to
1780 refs to unconstrained commons.
1782 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1785 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1786 look through SAVE_EXPRs with non-side-effects argument. Adjust
1788 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1791 2018-04-06 Richard Biener <rguenther@suse.de>
1794 * alias.c (find_base_term): New wrapper around find_base_term
1795 unwinding CSELIB_VAL_PTR changes.
1796 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1799 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1801 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1803 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1804 constant definitions.
1805 ("nop"): lr 0,0 -> nopr r0
1806 ("nop_lr0", "nop_lr1"): New insn definitions.
1808 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1810 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1811 NDS32_V3PUSH_AVAILABLE_P macro.
1813 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1814 Chung-Ju Wu <jasonwucj@gmail.com>
1816 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1817 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1818 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1819 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1820 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1821 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1822 * config/nds32/constraints.md: New constraints and checking for hard
1823 float configuration.
1824 * config/nds32/iterators.md: New mode iterator and attribute for hard
1825 float configuration.
1826 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1828 * config/nds32/nds32-fpu.md: New file.
1829 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1830 deal with hard float code generation.
1831 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1833 (abi_type, float_reg_number): New enum type.
1834 * config/nds32/nds32-predicates.c: New predicates for hard float.
1835 * config/nds32/nds32-protos.h: Declare functions for hard float.
1836 * config/nds32/nds32.c: Implementation for hard float configuration.
1837 * config/nds32/nds32.h: Definitions for hard float configuration.
1838 * config/nds32/nds32.md: Include hard float machine description and
1839 modify patterns for hard float configuration.
1840 * config/nds32/nds32.opt: New options for hard float configuration.
1841 * config/nds32/predicates.md: New predicates for hard float
1844 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1846 * common/config/nds32/nds32-common.c
1847 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1849 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1852 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1853 CONSTRUCTOR_ELT (ctor, ...)->value.
1855 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1858 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1860 2018-04-05 Tom de Vries <tom@codesourcery.com>
1863 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1866 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1867 Kito Cheng <kito.cheng@gmail.com>
1869 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1870 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1871 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1872 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1874 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1875 Kito Cheng <kito.cheng@gmail.com>
1877 * config/nds32/constraints.md (Ufe): New memory constraint.
1878 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1879 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1880 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1882 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1883 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1885 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config/nds32/nds32.md: Use optimize_size in the condition for
1888 alu-shift instructions.
1890 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1892 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1894 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1896 * config/nds32/nds32.md (negsi2): Refine pattern.
1898 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1899 Chung-Ju Wu <jasonwucj@gmail.com>
1901 * config/nds32/iterators.md (shift_rotate): New code iterator.
1902 (shift): New code attribute.
1903 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1904 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1905 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1906 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1907 bit-wise operations.
1908 (andsi3, *andsi3): Ditto.
1909 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1910 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1911 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1912 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1913 nds32_ior_operand, nds32_xor_operand): New predicates.
1915 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1917 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1918 (addsi3, subsi3): ... this.
1920 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1922 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1924 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1926 * config/nds32/nds32.md: Adjust indention.
1928 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1930 * config/nds32/nds32.md (feature): New attribute.
1932 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1934 * config/nds32/nds32.md (subtype): New attribute.
1936 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1939 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1940 expansion to perform a bitwise AND of the argument followed by a
1941 boolean negation of the result.
1943 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1945 PR rtl-optimization/84878
1946 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1947 the basic block. Assert the use reference is not artificial and that
1948 it has an associated insn.
1950 2018-04-04 Michael Matz <matz@suse.de>
1952 * builtins.c (compute_objsize): Pass correct operand
1953 to array_at_struct_end_p.
1955 2018-04-04 Richard Biener <rguenther@suse.de>
1958 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1959 from contexts for DINFO_LEVEL_TERSE and below.
1961 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1963 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1964 resiter_operand condition.
1965 * config/nds32/nds32.md (*move<mode>): Ditto.
1967 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1968 Monk Chiang <sh.chiang04@gmail.com>
1970 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1972 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1974 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1976 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1977 Kito Cheng <kito.cheng@gmail.com>
1979 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1980 nds32_cond_code_str, output_cond_branch,
1981 output_cond_branch_compare_zero, nds32_expand_cbranch,
1982 nds32_expand_cstore, nds32_expand_movcc,
1983 nds32_output_cbranchsi4_equality_zero,
1984 nds32_output_cbranchsi4_equality_reg,
1985 nds32_output_cbranchsi4_equality_reg_or_const_int,
1986 nds32_output_cbranchsi4_greater_less_zero: New functions.
1987 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1988 nds32_expand_cstore, nds32_expand_movcc,
1989 nds32_output_cbranchsi4_equality_zero,
1990 nds32_output_cbranchsi4_equality_reg,
1991 nds32_output_cbranchsi4_equality_reg_or_const_int,
1992 nds32_output_cbranchsi4_greater_less_zero): Declare.
1993 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1994 nds32_rimm11s_operand): New predicates.
1995 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1996 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1999 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2001 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2002 * config/nds32/nds32.md: Ditto.
2003 * config/nds32/pipelines.md: Ditto.
2005 2018-04-04 Richard Biener <rguenther@suse.de>
2007 PR tree-optimization/85168
2008 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2009 propagating abnormals.
2011 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2013 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2015 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2016 Kito Cheng <kito.cheng@gmail.com>
2018 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2019 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2020 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2021 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2022 * config/nds32/nds32.md (sibcall_internal): New.
2023 (sibcall_register): Remove.
2024 (sibcall_immediate): Remove.
2025 (sibcall_value_internal): New.
2026 (sibcall_value_register): Remove.
2027 (sibcall_value_immediate): Remove.
2028 * config/nds32/predicates.md (nds32_general_register_operand): New.
2029 (nds32_call_address_operand): New.
2031 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2033 PR rtl-optimization/85167
2034 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2035 bb_defs if *split_p, instead preinitialize it to NULL.
2037 PR tree-optimization/85156
2038 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2039 evaluating the argument multiple times.
2041 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2043 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2045 (_mm_cvtpd_ps): Likewise.
2046 (_mm_cvttpd_epi32): Likewise.
2047 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2048 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2049 vector, pixel, and bool following altivec.h include.
2051 2018-04-03 Martin Sebor <msebor@redhat.com>
2053 * doc/extend.texi (Common Function Attributes): Clarify.
2054 (const attribute): Likewise.
2055 (pure attribute): Likewise.
2057 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2060 * config/i386/i386.c (ix86_expand_vector_set): Use
2061 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2063 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2065 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2066 instructions when changing rounding bits to preserve precision bits
2067 in the x87 control word.
2069 2018-04-03 Martin Liska <mliska@suse.cz>
2071 PR tree-optimization/82491
2072 * rtl.h (strip_offset_and_add): Replace += suboffset with
2073 poly_uint64 () + suboffset.
2075 2018-03-29 Martin Liska <mliska@suse.cz>
2076 Martin Jambor <mjambor@suse.cz>
2079 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2080 param_type is not an integral or pointer type.
2082 2018-04-03 Richard Biener <rguenther@suse.de>
2084 * sese.h (recompute_all_dominators): Remove.
2086 2018-04-02 Martin Sebor <msebor@redhat.com>
2088 * doc/invoke.texi (-Wrestrict): Fix typos.
2090 2018-04-02 Jim Wilson <jimw@sifive.com>
2092 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2093 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2094 (<optab>di3, <optab>si3_extend): Likewise.
2095 (<optab>si3_mask, <optab>si3_mask_1): New.
2096 (<optab>di3_mask, <optab>di3_mask_1): New.
2097 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2098 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2099 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2101 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2103 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2106 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2108 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2109 (nds32_canonicalize_comparison): New function.
2111 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2112 Kito Cheng <kito.cheng@gmail.com>
2113 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2115 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2116 * config/nds32/constants.md (unspec_volatile_element): Add
2117 UNSPEC_VOLATILE_RELAX_GROUP.
2118 * config/nds32/nds32-relax-opt.c: New file.
2119 * config/nds32/nds32-predicates.c
2120 (nds32_symbol_load_store_p): New function.
2121 * config/nds32/nds32-protos.h
2122 (nds32_symbol_load_store_p): Declare function.
2123 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2124 * config/nds32/nds32.c
2125 (nds32_register_pass): New function to register pass.
2126 (nds32_register_passes): New function to register passes.
2127 * config/nds32/nds32.md (relax_group): New pattern.
2128 * config/nds32/nds32.opt (mrelax-hint): New option.
2129 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2131 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2133 * config/nds32/t-nds32: Modify files dependency.
2135 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2137 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2138 (PROFILE_HOOK): Define its implementation.
2140 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2142 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2143 type and 32-bit size.
2145 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2148 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2149 (V_128_256): New mode iterator.
2150 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2151 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2152 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2154 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2155 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2157 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2160 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2161 NaN inputs correctly.
2163 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2166 * config/rs6000/vsx.md (??r): New mode attribute.
2167 (*vsx_mov<mode>_64bit): Use it.
2168 (*vsx_mov<mode>_32bit): Likewise.
2170 2018-03-30 Martin Sebor <msebor@redhat.com>
2172 PR tree-optimization/84818
2173 * builtins.c (check_access): Use warning_n.
2175 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2178 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2180 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2183 2018-03-30 Julia Koval <julia.koval@intel.com>
2186 * x86-tune.def (movx, partial_reg_dependency): Enable for
2189 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2192 * lra-constraints.c (process_alt_operands): Move setting
2193 this_alternative_matches below.
2195 2018-03-29 Martin Liska <mliska@suse.cz>
2198 * doc/invoke.texi: Document how LTO works with debug info.
2199 Describe auto-load support of binutils. Mention 'x86-64'
2200 as valid option value of -march option.
2202 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2204 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2207 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2208 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2209 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2212 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2215 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2216 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2217 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2218 for RS6000_BTM_POWERPC64.
2219 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2220 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2221 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2226 2018-03-28 Carl Love <cel@us.ibm.com>
2229 2017-09-27 Carl Love <cel@us.ibm.com>
2231 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2232 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2233 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2236 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2238 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2239 instead of __vector bool.
2240 (_mm_max_pu8): Likewise.
2241 (_mm_min_pi16): Likewise.
2243 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2246 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2247 (DIVWEUO): Likewise.
2249 (DIVDEUO): Likewise.
2250 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2251 DIVWEUO and DIVDEUO.
2252 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2253 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2254 (div_extend): Likewise.
2255 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2257 (__builtin_divweuo): Likewise.
2258 (__builtin_divdeo): Likewise.
2259 (__builtin_divdeuo): Likewise.
2261 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2264 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2265 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2267 PR tree-optimization/82004
2268 * gimple-match-head.c (optimize_pow_to_exp): New function.
2269 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2270 Don't fold to exp if optimize_pow_to_exp is false.
2272 2018-03-28 Martin Liska <mliska@suse.cz>
2275 * calls.c (initialize_argument_information): Fix trailing space.
2276 * common.opt: Fix typo and provide better explanation for
2277 -fsanitize-coverage option.
2278 * config/i386/i386.opt: Fix typo.
2280 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2281 Martin Liska <mliska@suse.cz>
2284 * gimplify.c (asan_poison_variable): Don't do the check for
2285 gimplify_omp_ctxp here.
2286 (gimplify_decl_expr): Do it here.
2287 (gimplify_target_expr): Likewise.
2289 2018-03-28 Martin Liska <mliska@suse.cz>
2292 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2293 chkp_type_bounds_count if MPX is not enabled.
2295 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2297 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2299 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2302 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2303 function to create the function decl for complex long double
2304 multiply and divide for -mabi=ieeelongdouble.
2305 (init_float128_ieee): Call it.
2307 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2310 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2311 -fcf-protection=branch -mibt.
2312 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2314 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2317 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2319 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2322 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2323 extern array declarations.
2325 2018-03-27 Richard Biener <rguenther@suse.de>
2328 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2329 explicit single_use checks.
2331 2018-03-27 Richard Biener <rguenther@suse.de>
2333 PR tree-optimization/85082
2334 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2337 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2339 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2340 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2341 Turn on fasynchronous-unwind-tables and funwind-tables.
2343 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2346 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2347 (*bmi_blsr_<mode>_ccz): Ditto.
2349 2018-03-26 Tom de Vries <tom@codesourcery.com>
2351 PR tree-optimization/85063
2352 * omp-general.c (offloading_function_p): New function. Factor out
2354 * omp-offload.c (pass_omp_target_link::gate): ... here.
2355 * omp-general.h (offloading_function_p): Declare.
2356 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2357 with attribute omp declare target for offloading functions.
2359 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2361 PR tree-optimization/84005
2362 * tree-data-ref.h (get_base_for_alignment): Declare.
2363 * tree-data-ref.c (get_base_for_alignment_1): New function.
2364 (get_base_for_alignment): Likewise.
2365 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2366 get_base_for_alignment to find a suitable base object, instead
2367 of always using drb->base_address.
2369 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2372 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2373 known size by default.
2375 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2378 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2379 and non BLKmode operands.
2381 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2384 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2385 Clean up attributes.
2387 2018-03-23 Richard Biener <rguenther@suse.de>
2390 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2391 we are going to emit early debug for LTO.
2393 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2396 * function.c (match_asm_constraints_1): Don't optimize if input
2397 doesn't satisfy general_operand predicate for output's mode.
2400 * alias.c (write_dependence_p): Don't require for x_canonicalized
2401 non-VOIDmode if x has VOIDmode.
2404 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2405 just don't try to optimize it rather than assert it never happens.
2407 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2409 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2410 macro expansions for definition of ST_INTERNAL_<mode> and
2411 LD_INTERNAL_<mode> builtins.
2412 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2414 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2416 (altivec_expand_st_builtin): Likewise.
2417 (altivec_expand_builtin): Remove calls to deleted functions.
2418 (rs6000_address_for_altivec): Delete this function.
2419 * config/rs6000/vector.md: Remove expands for
2420 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2422 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2425 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2426 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2427 re-computing once computed.
2428 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2429 (arm_init_machine_status): Initialize
2430 machine->static_chain_stack_bytes.
2432 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2435 * doc/extend.texi: Add four new prototypes for vec_ld.
2436 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2437 definitions for more logical presentation.
2438 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2439 entries for V1TI variants of __builtin_altivec_ld builtin.
2440 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2441 handling of V1TI variant of LVX icode pattern.
2442 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2443 (rs6000_gimple_fold_builtin): Likewise.
2444 (altivec_init_builtins): Add code to define
2445 __builtin_altivec_lvx_v1ti function.
2447 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2450 * function.c (match_asm_constraints_1): Don't do the optimization
2451 if input isn't a REG, SUBREG, MEM or constant.
2453 2018-03-22 Tom de Vries <tom@codesourcery.com>
2455 PR tree-optimization/84956
2456 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2457 bb_has_abnormal_pred.
2459 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2462 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2463 DECL_INITIAL (decl) to decl at the end.
2464 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2467 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2469 * doc/extend.texi (__builtin_tgmath): Document when complex
2470 integer types are treated as _Complex _Float64.
2472 2018-03-21 Tom de Vries <tom@codesourcery.com>
2474 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2476 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2478 PR tree-optimization/84960
2479 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2480 if it is ENTRY block, move them into single succ of ENTRY in that case.
2482 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2484 PR tree-optimization/84811
2485 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2486 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2487 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2488 (known_subrange_p): Update accordingly. Cast each value involved
2489 in the size comparison, rather than casting the result of the
2492 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2494 PR tree-optimization/84982
2495 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2496 by flipping the least significant bit rather than all bits from
2497 bitpos to bitpos + bitsize - 1.
2499 2018-03-21 Nathan Sidwell <nathan@acm.org>
2501 * doc/extend.texi (Deprecated Features): Remove mention of
2502 long-deleted deprecations.
2504 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2507 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2508 * configure: Regenerate.
2510 2018-03-21 Tom de Vries <tom@codesourcery.com>
2512 PR tree-optimization/83126
2513 * tree-parloops.c (num_phis): New function.
2514 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2516 2018-03-21 Nathan Sidwell <nathan@acm.org>
2518 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2519 mention anon-struct/union members and trailing attributes.
2521 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2523 PR tree-optimization/84969
2524 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2525 builtin memset partitions if they set different rhs values.
2527 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2529 PR rtl-optimization/84989
2530 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2531 VEC_DUPLICATE with scalar result mode.
2533 2018-03-21 Martin Liska <mliska@suse.cz>
2536 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2537 not intended return statement.
2539 2018-03-21 Martin Liska <mliska@suse.cz>
2542 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2543 (chkp_find_bound_slots_1): Limit number of iterations.
2545 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2548 * Minor grammar fixes for x86 options.
2550 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2553 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2554 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2556 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2559 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2560 (altivec_lvx_<mode>_1op): Likewise.
2561 (altivec_stvx_<mode>_2op): Likewise.
2562 (altivec_stvx_<mode>_1op): Likewise.
2563 (altivec_lvx_<VM2:mode>): New define_expand.
2564 (altivec_stvx_<VM2:mode>): Likewise.
2565 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2566 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2567 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2568 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2569 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2570 (rs6000_gen_lvx): Likewise.
2571 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2572 (altivec_expand_stv_builtin): Likewise.
2573 (altivec_expand_builtin): Likewise.
2574 * config/rs6000/vector.md: Likewise.
2576 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2579 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2582 2018-03-20 Richard Biener <rguenther@suse.de>
2585 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2586 sign-conversions as zero, fall back to standard scalar_stmt
2589 2018-03-20 Martin Liska <mliska@suse.cz>
2592 * predict.c (rebuild_frequencies): Handle case when we have
2593 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2595 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2598 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2599 flag_section_anchors.
2600 * varasm.c (use_blocks_for_decl_p): Remove hack for
2601 dw2_force_const_mem.
2604 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2606 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2607 be created, use lowpart_subreg of operands[0] rather than operands[0]
2609 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2610 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2611 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2612 and n constraint instead of aarch64_shift_imm_di and Usd.
2613 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2614 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2616 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2619 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2620 to favor GPR over NEON registers.
2621 (<shift>di3_neon): Likewise.
2623 2018-03-20 Tom de Vries <tom@codesourcery.com>
2626 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2627 (nvptx_process_pars): Emit bar.sync asap and alap.
2629 2018-03-20 Tom de Vries <tom@codesourcery.com>
2632 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2633 seen_label if seen_label is already set.
2635 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2638 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2639 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2640 Use 1U instead of 1. Formatting fixes.
2643 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2644 instead of TREE_TYPE (s1) for the return value.
2646 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2648 PR tree-optimization/84946
2649 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2650 bitsize + bitsize in poly_uint64 rather than poly_int64.
2653 * dwarf2asm.c: Include fold-const.c.
2654 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2655 of decl rather than decl itself.
2657 PR rtl-optimization/84643
2658 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2660 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2663 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2664 calling assemble_variable.
2666 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2669 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2670 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2672 2018-03-19 Jim Wilson <jimw@sifive.com>
2675 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2676 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2677 (riscv_first_stack_step): Likewise.
2678 (riscv_option_override): Use STACK_BOUNDARY instead of
2680 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2682 (BIGGEST_ALIGNMENT): Set to 128.
2683 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2684 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2687 2018-03-19 Richard Biener <rguenther@suse.de>
2689 PR tree-optimization/84933
2690 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2691 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2693 2018-03-19 Richard Biener <rguenther@suse.de>
2695 PR tree-optimization/84859
2696 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2697 (cond_if_else_store_replacement): Perform sinking operation on
2698 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2699 Generalize what a BB with a single eligible store is.
2701 2018-03-19 Richard Biener <rguenther@suse.de>
2703 PR tree-optimization/84929
2704 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2705 chrec_is_positive against non-chrec arg.
2707 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2710 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2712 2018-03-18 Martin Liska <mliska@suse.cz>
2714 PR rtl-optimization/84635
2715 * regrename.c (build_def_use): Use matches_mode only when
2718 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2720 PR tree-optimization/84913
2721 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2722 vectorize chains of COND_EXPRs.
2724 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2726 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2728 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2730 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2732 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2734 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2736 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2737 Kito Cheng <kito.cheng@gmail.com>
2739 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2740 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2741 (nds32_adjust_reg_alloc_order): New function.
2742 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2744 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2746 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2747 nds32_print_operand, nds32_print_operand_address): Use
2748 HOST_WIDE_INT_PRINT_DEC instead.
2750 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2752 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2754 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2757 * config/i386/i386.c (initial_ix86_tune_features,
2758 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2760 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2761 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2762 rather than 1u << ix86_tune. Formatting fix.
2763 (ix86_option_override_internal): Change ix86_arch_mask from
2764 unsigned int to unsigned HOST_WIDE_INT, initialize to
2765 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2766 (ix86_function_specific_restore): Likewise.
2768 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2771 * postreload.c (reload_combine_recognize_pattern): Perform
2772 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2773 truncate_int_for_mode the result for the destination's mode.
2776 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2777 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2779 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2782 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2785 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2786 regno_allocno_class_array and sorted_pseudos.
2787 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2788 insns where regno is used.
2790 2018-03-16 Martin Liska <mliska@suse.cz>
2793 * multiple_target.c (create_dispatcher_calls): Redirect
2794 reference in the symbol table.
2796 2018-03-16 Martin Liska <mliska@suse.cz>
2799 * multiple_target.c (create_dispatcher_calls): Redirect also
2802 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2806 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2807 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2808 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2810 2018-03-16 Julia Koval <julia.koval@intel.com>
2812 * doc/invoke.texi (Skylake Server): Add CLWB.
2813 Cannonlake): Remove CLWB.
2815 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2817 PR tree-optimization/84841
2818 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2820 (FLOAT_ONE_CONST_TYPE): Define.
2821 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2822 (sort_by_operand_rank): Put entries with higher constant_type last
2823 rather than first to match comments.
2825 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2827 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2830 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2833 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2834 check and use address of target always.
2836 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2839 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2840 (indirect_thunk_bnd_needed): Likewise.
2841 (indirect_thunks_used): Likewise.
2842 (indirect_thunks_bnd_used): Likewise.
2843 (indirect_return_needed): New.
2844 (indirect_return_bnd_needed): Likewise.
2845 (output_indirect_thunk_function): Add a bool argument for
2847 (output_indirect_thunk_function): Don't generate alias for
2848 function return thunk.
2849 (ix86_code_end): Call output_indirect_thunk_function to generate
2850 function return thunks.
2851 (ix86_output_function_return): Set indirect_return_bnd_needed
2852 and indirect_return_needed instead of indirect_thunk_bnd_needed
2853 and indirect_thunk_needed.
2855 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2857 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2858 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2859 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2861 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2862 Paul Hua <paul.hua.gm@gmail.com>
2865 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2867 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2869 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2870 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2873 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2876 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2877 instead of GET_MODE_SIZE when comparing Units.
2879 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2882 * varasm.c (hash_section): Return an unchangeble hash value
2883 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2884 Return !aarch64_can_use_per_function_literal_pools_p ().
2886 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2889 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2890 rather than address of cmode as last argument to prepare_cmp_insn.
2892 2018-03-15 Julia Koval <julia.koval@intel.com>
2894 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2895 F_AVX512VNNI, F_AVX512BITALG): New.
2897 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2900 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2901 insn for floating-point loads and stores.
2903 2018-03-14 Carl Love <cel@us.ibm.com>
2905 * config/rs6000/rs6000-c.c: Add macro definitions for
2906 ALTIVEC_BUILTIN_VEC_PERMXOR.
2907 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2908 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2909 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2910 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2912 * config/doc/extend.texi: Add prototypes for vec_permxor.
2914 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2917 * diagnostic-show-locus.c (class layout_point): Convert m_line
2918 from int to linenum_type.
2919 (line_span::comparator): Use linenum "compare" function when
2920 comparing line numbers.
2921 (test_line_span): New function.
2922 (layout_range::contains_point): Convert param "row" from int to
2924 (layout_range::intersects_line_p): Likewise.
2925 (layout::will_show_line_p): Likewise.
2926 (layout::print_source_line): Likewise.
2927 (layout::should_print_annotation_line_p): Likewise.
2928 (layout::print_annotation_line): Likewise.
2929 (layout::print_leading_fixits): Likewise.
2930 (layout::annotation_line_showed_range_p): Likewise.
2931 (struct line_corrections): Likewise for field m_row.
2932 (line_corrections::line_corrections): Likewise for param "row".
2933 (layout::print_trailing_fixits): Likewise.
2934 (layout::get_state_at_point): Likewise.
2935 (layout::get_x_bound_for_row): Likewise.
2936 (layout::print_line): Likewise.
2937 (diagnostic_show_locus): Likewise for locals "last_line" and
2939 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2940 * input.c (selftest::test_linenum_comparisons): New function.
2941 (selftest::input_c_tests): Call it.
2942 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2943 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2944 * selftest.h (ASSERT_GT): New macro.
2945 (ASSERT_GT_AT): New macro.
2946 (ASSERT_LT): New macro.
2947 (ASSERT_LT_AT): New macro.
2949 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2951 PR rtl-optimization/84780
2952 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2954 2018-03-14 Martin Liska <mliska@suse.cz>
2956 * tree.c (record_node_allocation_statistics): Use
2957 get_stats_node_kind.
2958 (get_stats_node_kind): New function extracted from
2959 record_node_allocation_statistics.
2960 (free_node): Use get_stats_node_kind.
2962 2018-03-14 Richard Biener <rguenther@suse.de>
2964 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2965 that the value-set of ANTIC_IN doesn't grow.
2968 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2970 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2971 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2973 2018-03-14 Julia Koval <julia.koval@intel.com>
2975 * config.gcc (icelake-client, icelake-server): New.
2977 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2978 (initial_ix86_arch_features): Ditto.
2979 (PTA_SKYLAKE): Add SGX.
2980 (PTA_ICELAKE): Remove.
2981 (PTA_ICELAKE_CLIENT): New.
2982 (PTA_ICELAKE_SERVER): New.
2983 (ix86_option_override_internal): Split up icelake on icelake client and
2985 (get_builtin_code_for_version): Ditto.
2986 (fold_builtin_cpu): Ditto.
2987 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2988 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2989 * config/i386/i386.h (processor_type): Ditto.
2990 * doc/invoke.texi: Ditto.
2992 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2995 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2996 INTEGER_CST offset, add it together with bitpos / 8 and
2997 sign extend based on POINTER_SIZE.
3001 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3004 * config/i386/constraints.md (Yc): New register constraint.
3005 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3006 Use Yc constraint for alternative 2 of operand 0. Remove
3007 preferred_for_speed attribute.
3009 2018-03-14 Richard Biener <rguenther@suse.de>
3011 PR tree-optimization/84830
3012 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3013 with the old one to avoid oscillations.
3015 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3018 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3020 (assign_by_spills): Return a flag of reload assignment failure.
3021 Do not process the reload assignment failures. Do not spill other
3022 reload pseudos if they has the same reg class. Update n if
3024 (lra_assign): Add a return arg. Set up from the result of
3025 assign_by_spills call.
3026 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3027 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3028 usage_insns if it is not NULL.
3029 (spill_hard_reg_in_range): New function.
3030 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3031 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3032 function prototypes.
3033 (lra_assign): Change prototype.
3034 * lra.c (lra): Add code to deal with fails by splitting hard reg
3037 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3039 * config/riscv/riscv.opt (mrelax): New option.
3040 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3041 "norelax" when riscv_mrelax is disabled.
3042 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3044 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3047 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3048 reassociation for int modes.
3050 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3052 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3053 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3055 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3056 * config/aarch64/aarch64-sve.md
3057 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3058 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3059 (*extend<mode><Vwide>2): Rename to...
3060 (aarch64_sve_extend<mode><Vwide>2): ...this.
3061 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3062 renaming the old pattern to...
3063 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3065 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3066 define_expand, renaming the old pattern to...
3067 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3068 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3069 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3070 account when deciding which SVE instruction the optab should use.
3071 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3073 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3075 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3076 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3077 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3078 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3079 (tlsdesc_small_<mode>): Turn a define_expand and use
3080 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3081 (tlsdesc_small_advsimd_<mode>): ...this.
3082 (tlsdesc_small_sve_<mode>): New pattern.
3084 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3086 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3087 (UNSPEC_UMUL_HIGHPART): New constants.
3088 (MUL_HIGHPART): New int iteraor.
3089 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3090 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3092 (*<su>mul<mode>3_highpart): New define_insn.
3094 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3097 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3100 2018-03-13 Martin Liska <mliska@suse.cz>
3103 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3105 (sem_item_optimizer::~sem_item_optimizer): Release it.
3106 (sem_item_optimizer::merge_classes): Register variable aliases.
3107 (sem_item_optimizer::fixup_pt_set): New function.
3108 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3109 * ipa-icf.h: Declare new variables and functions.
3111 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3114 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3115 integer_pow2p@2 and test integer_pow2p in condition.
3116 (A < 0 ? C : 0): Similarly for @1.
3119 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3120 characters starting at p contain '\0' character, don't look beyond
3124 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3125 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3128 * reg-stack.c (change_stack): Change update_end var from int to
3129 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3130 also call set_block_for_insn on the newly added insns and rescan.
3133 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3134 on the last operand.
3137 * tree.c (stabilize_reference_1): Return save_expr (e) for
3138 STATEMENT_LIST even if it doesn't have side-effects.
3140 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3142 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3144 2018-03-12 Renlin Li <renlin.li@arm.com>
3146 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3147 aarch64_output_scalar_simd_mov_immediate.
3149 2018-03-12 Martin Sebor <msebor@redhat.com>
3151 PR tree-optimization/83456
3152 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3153 for perfectly overlapping calls to memcpy.
3154 (gimple_fold_builtin_memory_chk): Same.
3155 (gimple_fold_builtin_strcpy): Handle no-warning.
3156 (gimple_fold_builtin_stxcpy_chk): Same.
3157 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3159 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3161 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3162 parameter. Use it for SFmode.
3163 (rs6000_function_arg_advance_1): Adjust.
3164 (rs6000_function_arg): Adjust.
3165 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3167 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3169 PR rtl-optimization/84169
3170 PR rtl-optimization/84780
3171 * combine.c (can_combine_p): Check for a 2-insn combination whether
3172 the destination register is used between the two insns, too.
3174 2018-03-12 Richard Biener <rguenther@suse.de>
3176 PR tree-optimization/84803
3177 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3178 for refs DR analysis didn't process.
3180 2018-03-12 Richard Biener <rguenther@suse.de>
3182 PR tree-optimization/84777
3183 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3184 force-vectorize loops ignore whether we are optimizing for size.
3186 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3188 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3189 (TARGET_MD_ASM_ADJUST): Define.
3191 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3192 Kito Cheng <kito.cheng@gmail.com>
3193 Chung-Ju Wu <jasonwucj@gmail.com>
3195 * config/nds32/nds32.c (nds32_compute_stack_frame,
3196 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3197 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3198 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3199 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3200 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3201 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3202 * config/nds32/nds32.md (prologue, epilogue): Use macro
3203 NDS32_V3PUSH_AVAILABLE_P to do checking.
3205 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3208 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3209 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3210 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3211 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3212 addition of most attributes on !orig_type_die or the attribute not
3213 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3215 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3216 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3220 * config/nds32/nds32.opt (mvh): New option.
3222 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3223 Chung-Ju Wu <jasonwucj@gmail.com>
3225 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3227 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3228 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3231 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3232 Chung-Ju Wu <jasonwucj@gmail.com>
3234 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3236 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3237 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3239 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3240 Kito Cheng <kito.cheng@gmail.com>
3241 Chung-Ju Wu <jasonwucj@gmail.com>
3243 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3244 UNSPEC_FFMISM and UNSPEC_FLMISM.
3245 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3246 for ffb, ffmism and flmism.
3247 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3248 (unspec_ffmism): Ditto.
3249 (unspec_flmism): Ditto.
3250 (nds32_expand_builtin_impl): Check if string extension is available.
3251 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3252 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3254 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3257 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3260 * lra-assigns.c (assign_by_spills): Return a flag of reload
3261 assignment failure. Do not process the reload assignment
3262 failures. Do not spill other reload pseudos if they has the same
3264 (lra_assign): Add a return arg. Set up from the result of
3265 assign_by_spills call.
3266 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3267 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3268 usage_insns if it is not NULL.
3269 (spill_hard_reg_in_range): New function.
3270 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3271 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3272 function prototypes.
3273 (lra_assign): Change prototype.
3274 * lra.c (lra): Add code to deal with fails by splitting hard reg
3277 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3280 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3282 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3285 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3286 (dw_val_node): Add val_symbolic_view.
3287 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3288 (symview_upper_bound): New.
3289 (new_line_info_table): Initialize symviews_since_reset.
3290 (dwarf2out_source_line): Count symviews_since_reset and set
3291 symview_upper_bound.
3292 (dw_val_equal_p): Handle symview.
3293 (add_AT_symview): New.
3294 (print_dw_val): Handle symview.
3295 (attr_checksum, attr_checksum_ordered): Likewise.
3296 (same_dw_val_p, size_of_die): Likewise.
3297 (value_format, output_die): Likewise.
3298 (add_high_low_attributes): Use add_AT_symview for entry_view.
3299 (dwarf2out_finish): Reset symview_upper_bound, clear
3302 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3305 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3306 Add strict argument and use it.
3307 (rs6000_split_multireg_move): Update for new strict argument.
3308 (mem_operand_gpr): Disallow all non-offsettable addresses.
3309 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3311 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3314 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3315 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3316 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3319 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3320 decl, use remap_type if we want to use the type.
3322 2018-03-09 Martin Sebor <msebor@redhat.com>
3324 PR tree-optimization/84526
3325 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3327 (builtin_access::generic_overlap): Be prepared to handle non-array
3330 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3332 PR rtl-optimization/84682
3333 * lra-constraints.c (process_address_1): Check is_address flag
3334 for address constraints.
3335 (process_alt_operands): Likewise.
3336 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3337 preprocess_constraints.
3338 * recog.h (preprocess_constraints): Add oploc parameter.
3340 * recog.c (preprocess_constraints): Test address_operand for
3341 CT_ADDRESS constraints.
3343 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3346 * lra-assigns.c (assign_by_spills): Return a flag of reload
3347 assignment failure. Do not process the reload assignment
3348 failures. Do not spill other reload pseudos if they has the same
3350 (lra_assign): Add a return arg. Set up from the result of
3351 assign_by_spills call.
3352 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3353 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3354 usage_insns if it is not NULL.
3355 (spill_hard_reg_in_range): New function.
3356 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3357 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3358 function prototypes.
3359 (lra_assign): Change prototype.
3360 * lra.c (lra): Add code to deal with fails by splitting hard reg
3363 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3366 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3367 Accept complain bool parameter. Only emit errors if it is true.
3368 (arm_parse_cpu_option_name): Likewise.
3369 (arm_target_thumb_only): Adjust callers of the above.
3370 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3371 prototype to take a default true bool parameter.
3372 (arm_parse_arch_option_name): Likewise.
3374 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3375 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3379 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3380 * configure: Regenerate.
3381 * configure.ac ("linker --version-script option"): New.
3382 ("linker soname option"): New.
3384 2018-03-09 Richard Biener <rguenther@suse.de>
3386 PR tree-optimization/84775
3387 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3388 immediate uses of predicate stmts and mark them modified.
3391 PR tree-optimization/84178
3392 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3394 (version_loop_for_if_conversion): Delay update_ssa call.
3395 (tree_if_conversion): Delay update_ssa until after predicate
3398 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3401 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3402 when the function accesses prior frames.
3404 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3407 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3408 gen_llsym, otherwise call maybe_gen_llsym.
3411 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3412 has ',' character inside of it.
3414 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3417 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3418 as clobbering CC_REGNUM.
3420 2018-03-08 Richard Biener <rguenther@suse.de>
3423 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3424 (follow_copies_to_constant): Do not follow SSA names registered
3427 2018-03-08 Richard Biener <rguenther@suse.de>
3429 PR tree-optimization/84178
3430 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3432 (version_loop_for_if_conversion): Delay update_ssa call.
3433 (tree_if_conversion): Delay update_ssa until after predicate
3436 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3438 PR tree-optimization/84178
3439 * tree-if-conv.c (release_bb_predicate): Remove the
3440 the assertion that the stmts have NULL use_ops.
3441 Discard the statements, asserting that they haven't
3442 yet been added to a BB.
3444 2018-03-08 Richard Biener <rguenther@suse.de>
3446 PR tree-optimization/84746
3447 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3448 (phi_translate): Pass in destination ANTIC_OUT set.
3449 (phi_translate_1): Likewise. For a simplified result lookup
3450 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3451 (phi_translate_set): Adjust.
3452 (do_pre_regular_insertion): Likewise.
3453 (do_pre_partial_partial_insertion): Likewise.
3455 2018-03-08 Martin Liska <mliska@suse.cz>
3457 PR gcov-profile/84735
3458 * doc/gcov.texi: Document usage of profile files.
3459 * gcov-io.h: Document changes in the format.
3461 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3465 * dwarf2out.c (struct dw_line_info_table): Update comments for
3467 (FORCE_RESET_NEXT_VIEW): New.
3468 (FORCE_RESETTING_VIEW_P): New.
3469 (RESETTING_VIEW_P): Check for -1 too.
3470 (ZERO_VIEW_P): Likewise.
3471 (new_line_info_table): Force-reset next view.
3472 (dwarf2out_begin_function): Likewise.
3473 (dwarf2out_source_line): Simplify zero_view_p initialization.
3474 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3475 view directly. Omit view when omitting .loc at line 0.
3477 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3479 PR tree-optimization/84740
3480 * tree-switch-conversion.c (process_switch): Call build_constructors
3481 only if info.phi_count is non-zero.
3483 PR tree-optimization/84739
3484 * tree-tailcall.c (find_tail_calls): Check call arguments against
3485 DECL_ARGUMENTS (current_function_decl) rather than
3486 DECL_ARGUMENTS (func) when checking for tail recursion.
3488 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3490 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3491 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3492 Volker Reichelt's entry and add entries for people that perform
3493 GCC fuzzy testing and report numerous bugs.
3495 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3498 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3499 readonly data in sdata, if that is disabled.
3500 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3501 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3502 -mreadonly-in-sdata option.
3504 2018-03-07 Martin Sebor <msebor@redhat.com>
3506 PR tree-optimization/84468
3507 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3508 basic block when looking for nul assignment.
3510 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3513 * except.h (output_function_exception_table): Adjust prototype.
3514 * except.c (output_function_exception_table): Remove FNNAME parameter
3515 and add SECTION parameter. Ouput one part of the table at a time.
3516 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3517 the first part of the exception table and emit unwind directives.
3518 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3519 (i386_pe_seh_cold_init): Likewise.
3520 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3521 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3522 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3523 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3524 * config/i386/winnt.c: Include except.h.
3525 (struct seh_frame_state): Add reg_offset, after_prologue and
3526 in_cold_section fields.
3527 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3528 (i386_pe_seh_cold_init): New function.
3529 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3530 to seh->in_cold_section.
3531 (seh_emit_push): Record the offset of the push.
3532 (seh_emit_save): Record the offet of the save.
3533 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3534 Test seh->after_prologue to disregard the epilogue.
3535 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3536 (i386_pe_end_cold_function): New function.
3538 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3541 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3542 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3545 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3547 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3548 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3551 * multiple_target.c: Include tree-inline.h and intl.h.
3552 (expand_target_clones): Diagnose and fail if node->definition and
3553 !tree_versionable_function_p (node->decl).
3555 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3557 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3559 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3560 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3561 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3563 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3566 * combine.c (try_combine): Use reg_or_subregno instead of handling
3567 just paradoxical SUBREGs and REGs.
3569 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3571 * config/arc/arc.c (arc_finalize_pic): Remove function.
3572 (arc_must_save_register): We use single base PIC register, remove
3573 checks to save/restore the PIC register.
3574 (arc_expand_prologue): Likewise.
3575 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3577 (arc_verify_short): Likewise.
3578 (arc_attr_type): Likewise.
3579 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3580 (walk_stores): Likewise.
3581 (arc_address_cost): Make it static.
3582 (arc_verify_short): Likewise.
3583 (branch_dest): Likewise.
3584 (arc_attr_type): Likewise.
3585 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3586 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3587 (arc_final_prescan_insn): Remove inserting the nops due to
3588 hardware hazards. It is done in reorg step.
3589 (insn_length_variant_t): Remove.
3590 (insn_length_parameters_t): Likewise.
3591 (arc_insn_length_parameters): Likewise.
3592 (arc_get_insn_variants): Likewise.
3593 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3595 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3598 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3601 PR tree-optimization/84687
3602 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3604 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3606 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3608 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3609 Rename to ppc_speculation_barrier.
3610 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3611 __builtin_ppc_speculation_barrier.
3613 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3616 * combine.c (combine_simplify_rtx): Don't try to simplify if
3617 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3620 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3622 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3623 to 32 bytes when compiling for POWER9.
3625 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3628 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3629 regparm >= 3 with no arg reg available also for calls with
3630 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3633 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3635 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3637 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3640 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3642 2018-03-05 Richard Biener <rguenther@suse.de>
3644 PR tree-optimization/84486
3645 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3646 When inserting a __builtin_assume_aligned call set the LHS
3647 SSA name alignment info accordingly.
3649 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3651 PR tree-optimization/84114
3652 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3653 Avoid reassociation of FLOAT_MODE addition.
3655 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3657 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3658 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3659 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3660 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3661 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3662 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3663 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3665 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3666 __builtin_ia32_wbinvd): New builtins.
3667 (SPECIAL_ARGS2): New.
3668 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3669 (SPECIAL_ARGS2): New.
3670 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3671 (ix86_valid_target_attribute_inner_p): Ditto.
3672 (ix86_init_mmx_sse_builtins): Add special_args2.
3673 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3674 TARGET_WBNOINVD_P): New.
3675 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3676 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3677 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3678 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3679 * config/i386/pconfigintrin.h: New file.
3680 * config/i386/wbnoinvdintrin.h: Ditto.
3681 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3682 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3684 2018-03-05 Richard Biener <rguenther@suse.de>
3686 PR tree-optimization/84670
3687 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3689 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3690 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3691 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3692 doesn't grow if all successors (recursively) were visited at least
3695 2018-03-05 Richard Biener <rguenther@suse.de>
3697 PR tree-optimization/84650
3698 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3699 if executed in the loop pipeline.
3701 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3703 * doc/configfiles.texi (Configuration Files): Move info about
3704 conditionalizing $target-protos.h to...
3705 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3706 differs from $target-protos.h.
3708 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3709 Chung-Ju Wu <jasonwucj@gmail.com>
3711 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3712 * config/nds32/nds32-multiple.md (setmemsi): Define.
3713 * config/nds32/nds32-memory-manipulation.c
3714 (nds32_gen_dup_4_byte_to_word_value): New.
3715 (emit_setmem_word_loop): New.
3716 (emit_setmem_byte_loop): New.
3717 (nds32_expand_setmem_loop): New.
3718 (nds32_expand_setmem_loop_v3m): New.
3719 (nds32_expand_setmem_unroll): New.
3720 (nds32_expand_setmem): New.
3722 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3723 Chung-Ju Wu <jasonwucj@gmail.com>
3725 * config/nds32/nds32-memory-manipulation.c
3726 (nds32_emit_load_store): New.
3727 (nds32_emit_post_inc_load_store): New.
3728 (nds32_emit_mem_move): New.
3729 (nds32_emit_mem_move_block): New.
3730 (nds32_expand_movmemsi_loop_unknown_size): New.
3731 (nds32_expand_movmemsi_loop_known_size): New.
3732 (nds32_expand_movmemsi_loop): New.
3733 (nds32_expand_movmemsi_unroll): New.
3734 (nds32_expand_movmemqi): Rename ...
3735 (nds32_expand_movmemsi): ... to this.
3736 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3737 (movmemsi): ... to this.
3738 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3739 (nds32_expand_movmemsi): ... to this.
3741 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3742 Monk Chiang <sh.chiang04@gmail.com>
3743 Chung-Ju Wu <jasonwucj@gmail.com>
3745 * config/nds32/nds32-protos.h
3746 (nds32_expand_load_multiple): New arguments.
3747 (nds32_expand_store_multiple): Ditto.
3748 (nds32_valid_multiple_load_store): Rename ...
3749 (nds32_valid_multiple_load_store_p): ... to this.
3750 * config/nds32/nds32-memory-manipulation.c
3751 (nds32_expand_load_multiple): Refine implementation.
3752 (nds32_expand_store_multiple): Ditto.
3753 * config/nds32/nds32-multiple.md
3754 (load_multiple): Update nds32_expand_load_multiple interface.
3755 (store_multiple): Update nds32_expand_store_multiple interface.
3756 * config/nds32/nds32-predicates.c
3757 (nds32_valid_multiple_load_store): Rename ...
3758 (nds32_valid_multiple_load_store_p): ... to this and refine
3760 * config/nds32/predicates.md
3761 (nds32_load_multiple_and_update_address_operation): New predicate.
3762 (nds32_store_multiple_and_update_address_operation): New predicate.
3764 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3765 Chung-Ju Wu <jasonwucj@gmail.com>
3767 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3768 (combo): New attribute.
3769 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3771 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3773 * config/nds32/nds32.opt: Change -mcmodel= default value.
3775 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3776 Monk Chiang <sh.chiang04@gmail.com>
3777 Chung-Ju Wu <jasonwucj@gmail.com>
3779 * config/nds32/constants.md (unspec_element): New enum.
3780 * config/nds32/constraints.md (Umw): New constraint.
3781 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3782 * config/nds32/nds32-intrinsic.md: Likewise.
3783 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3784 (nds32_valid_smw_lwm_base_p): New.
3785 (nds32_output_smw_single_word): New.
3786 (nds32_output_lmw_single_word): New.
3787 (nds32_expand_unaligned_load): New.
3788 (nds32_expand_unaligned_store): New.
3789 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3790 (nds32_output_smw_single_word): Declare.
3791 (nds32_output_lmw_single_word): Declare.
3792 (nds32_expand_unaligned_load): Declare.
3793 (nds32_expand_unaligned_store): Declare.
3794 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3795 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3796 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3797 NDS32_BUILTIN_UASTORE_DW.
3798 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3801 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3802 Kito Cheng <kito.cheng@gmail.com>
3803 Chung-Ju Wu <jasonwucj@gmail.com>
3805 * config/nds32/nds32-intrinsic.c
3806 (nds32_expand_builtin_null_ftype_reg): Delete.
3807 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3808 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3809 (nds32_read_argument): New.
3810 (nds32_legitimize_target): Ditto.
3811 (nds32_legitimize_argument): Ditto.
3812 (nds32_check_constant_argument): Ditto.
3813 (nds32_expand_unop_builtin): Ditto.
3814 (nds32_expand_unopimm_builtin): Ditto.
3815 (nds32_expand_binop_builtin): Ditto.
3816 (nds32_builtin_decl_impl): Ditto.
3817 (builtin_description): Ditto.
3818 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3819 (nds32_init_builtins_impl): Ditto.
3820 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3821 (nds32_builtin_decl): New.
3822 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3823 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3825 2018-03-02 Jeff Law <law@redhat.com>
3827 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3828 (redundant_insn, fill_simple_delay_slots): Likewise.
3829 (fill_slots_from_thread): Likewise.
3830 * resource.c (mark_referenced_resources): Likewise.
3831 (mark_set_resources, find_dead_or_set_registers): Likewise.
3833 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3835 * substring-locations.h (format_warning_va): Formatting fix for
3837 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3839 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3840 * substring-locations.c: Include intl.h.
3841 (format_warning_va): Turned into small wrapper around
3842 format_warning_n_va, renamed to ...
3843 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3844 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3846 (format_warning_at_substring_n): New function.
3847 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3848 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3849 format_warning_at_substring with just a shorter name instead of
3850 const function pointer.
3851 (fmtwarn_n): New function.
3852 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3853 appropriate, get rid of all the fmtstr temporaries, move conditionals
3854 with G_() wrapped string literals directly into fmtwarn arguments,
3855 cast dir.len to (int), formatting fixes.
3857 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3859 * doc/invoke.texi: Remove "Cilk Plus" references.
3861 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3862 Richard Biener <rguenther@suse.de>
3865 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3866 for error or warning attributes if CALL_FROM_THUNK_P is set.
3869 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3872 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3873 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3876 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3877 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3879 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3880 instead of -1U in last predictors element's probability member.
3882 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3885 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3886 arguments if they are comparable.
3888 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3890 PR tree-optimization/84634
3891 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3892 masks and masked_loop_p with a single loop_masks, making sure it's
3893 null for bb vectorization.
3895 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3897 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3898 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3899 loop->force_vectorize to check whether there is no alias.
3901 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3904 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3906 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3908 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3909 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3910 instead of a loop around prev_real_insn.
3911 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3915 * config/i386/i386.c (ix86_print_operand): Use conditional
3916 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3919 2018-03-02 Richard Biener <rguenther@suse.de>
3921 PR tree-optimization/84427
3922 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3923 (bitmap_set_subtract_values): Rewrite to handle multiple
3926 (prune_clobbered_mems): Likewise.
3927 (phi_translate): Take edge instead of pred/phiblock.
3928 (phi_translate_1): Likewise.
3929 (phi_translate_set): Likewise. Insert all translated
3930 exprs for a value into the set, keeping possibly multiple
3931 expressions per value.
3932 (compute_antic_aux): Adjust for phi_translate changes.
3933 When intersecting union the expressions and prune those
3934 not in the final value set, keeping possibly multiple
3935 expressions per value. Do not use value-insertion
3936 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3937 all expressions. Add verification that the value-sets
3938 only shrink during iteration.
3939 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3940 (do_pre_regular_insertion): Likewise.
3941 (do_pre_partial_partial_insertion): Likewise.
3943 2018-03-02 Richard Biener <rguenther@suse.de>
3946 * config/darwin.c (saved_debug_info_level): New static global.
3947 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3948 (darwin_asm_lto_end): Restore debug info generation settings.
3950 2018-03-01 Martin Liska <mliska@suse.cz>
3953 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3956 2018-03-01 Richard Biener <rguenther@suse.de>
3959 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3960 type annotation with LTO when debug was disabled at compile-time.
3962 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3964 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3966 (mips_final_postscan_insn): Likewise.
3968 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3970 PR rtl-optimization/84528
3971 * alias.c (init_alias_target): Add commentary.
3972 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3973 a unique base value if the frame pointer is not eliminated
3974 to the stack pointer.
3976 2018-03-01 Tom de Vries <tom@codesourcery.com>
3978 PR rtl-optimization/83327
3979 * lra-int.h (hard_regs_spilled_into): Declare.
3980 * lra.c (hard_regs_spilled_into): Define.
3981 (init_reg_info): Init hard_regs_spilled_into.
3982 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3983 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3984 (process_bb_lives): Handle hard_regs_spilled_into.
3985 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3986 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3988 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3990 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3991 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3992 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3993 * config/rs6000/aix72.h: New file.
3995 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3997 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3998 instead of warning_at with conditional singular and plural messages
4002 * stor-layout.c (update_alignment_for_field): For
4003 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4004 && !DECL_PACKED (field), do the alignment update, just use
4005 only desired_align instead of MAX (type_align, desired_align)
4007 (place_field): Don't do known_align < desired_align handling
4008 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4009 is non-NULL, instead do it after rli->prev_field handling and
4010 only if not within a bitfield word. For DECL_PACKED (field)
4011 use type_align of BITS_PER_UNIT.
4013 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4015 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4016 superfluous parentheses and trailing spaces.
4018 2018-02-28 Richard Biener <rguenther@suse.de>
4020 PR tree-optimization/84584
4021 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4022 SCoPs with fake exit edge.
4024 2018-02-28 Martin Liska <mliska@suse.cz>
4027 * timevar.c (timer::print): Fix format to properly print 100%
4030 2018-02-28 Richard Biener <rguenther@suse.de>
4033 * genmatch.c (capture_info::walk_match): Do not mark
4034 captured expressions without operands as expr_p given
4035 they act more like predicates and should be subject to
4036 "lost tail" side-effect preserving.
4038 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4040 PR rtl-optimization/81611
4041 * auto-inc-dec.c (attempt_change): Move dead note from
4042 mem_insn if it's the next use of regno
4043 (find_address): Take address use of reg holding
4044 non-incremented value. Add parm to limit search to the named
4046 (merge_in_block): Attempt to use a mem insn that is the next
4047 use of the original regno.
4049 2018-02-27 Martin Sebor <msebor@redhat.com>
4052 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4053 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4055 2018-02-27 Martin Sebor <msebor@redhat.com>
4057 PR translation/84207
4058 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4059 n argument to unsigned HOST_WIDE_INT.
4060 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4061 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4062 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4063 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4065 2018-02-27 Richard Biener <rguenther@suse.de>
4067 PR tree-optimization/84512
4068 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4069 Do not use the estimate returned from record_stmt_cost for
4070 the scalar iteration cost but sum properly using add_stmt_cost.
4072 2018-02-27 Richard Biener <rguenther@suse.de>
4074 PR tree-optimization/84466
4075 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4076 Adjust last change to less strictly validate use operands.
4078 2018-02-27 Martin Liska <mliska@suse.cz>
4080 PR gcov-profile/84548
4081 * gcov.c (process_file): Allow partial overlap and consider it
4082 also as group functions.
4083 (output_lines): Properly calculate range of lines for a group.
4085 2018-02-27 Martin Liska <mliska@suse.cz>
4087 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4088 'ggc' suffixes. Change first column width.
4089 (timer::print): Fix formatting of the column.
4091 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4093 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4094 preserve inline entry blocks for the sake of debug inline
4095 entry point markers alone.
4096 (remove_unused_locals): Suggest in comments a better place to
4097 force the preservation of inline entry blocks that are
4098 otherwise unused, but do not preserve them.
4100 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4102 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4104 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4107 * config/i386/constraints.md (Bs): Replace
4108 ix86_indirect_branch_register with
4109 TARGET_INDIRECT_BRANCH_REGISTER.
4111 * config/i386/i386.md (indirect_jump): Likewise.
4112 (tablejump): Likewise.
4113 (*sibcall_memory): Likewise.
4114 (*sibcall_value_memory): Likewise.
4115 Peepholes of indirect call and jump via memory: Likewise.
4116 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4117 (*sibcall_value_GOT_32): Likewise.
4118 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4119 (GOT_memory_operand): Likewise.
4120 (call_insn_operand): Likewise.
4121 (sibcall_insn_operand): Likewise.
4122 (GOT32_symbol_operand): Likewise.
4123 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4125 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4127 PR rtl-optimization/83496
4128 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4129 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4130 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4131 redundant insn, if any.
4132 (relax_delay_slots): Likewise.
4133 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4135 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4137 PR tree-optimization/83965
4138 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4139 that grouped statements are part of a reduction chain. Return
4140 true if the statement is not marked as a reduction itself but
4142 (vect_recog_dot_prod_pattern): Don't check whether the statement
4143 is part of a group here.
4144 (vect_recog_sad_pattern): Likewise.
4145 (vect_recog_widen_sum_pattern): Likewise.
4147 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4150 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4152 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4155 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4157 (ix86_output_indirect_function_return): New prototype.
4158 (ix86_split_simple_return_pop_internal): Likewise.
4159 * config/i386/i386.c (indirect_return_via_cx): New.
4160 (indirect_return_via_cx_bnd): Likewise.
4161 (indirect_thunk_name): Handle return va CX_REG.
4162 (output_indirect_thunk_function): Create alias for
4163 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4164 (ix86_output_indirect_jmp): Remove the bool argument.
4165 (ix86_output_indirect_function_return): New function.
4166 (ix86_split_simple_return_pop_internal): Likewise.
4167 * config/i386/i386.md (*indirect_jump): Don't pass false
4168 to ix86_output_indirect_jmp.
4169 (*tablejump_1): Likewise.
4170 (simple_return_pop_internal): Change it to define_insn_and_split.
4171 Call ix86_split_simple_return_pop_internal to split it for
4173 (simple_return_indirect_internal): Call
4174 ix86_output_indirect_function_return instead of
4175 ix86_output_indirect_jmp.
4177 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4180 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4181 memset and value initialization afterwards.
4183 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4185 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4187 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4190 * common/config/aarch64/aarch64-common.c
4191 (aarch_option_optimization_table[]): Switch
4192 off fomit-frame-pointer
4194 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4195 Chung-Ju Wu <jasonwucj@gmail.com>
4197 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4199 (store_multiple): Ditto.
4201 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4203 * config.gcc: Add --with-cpu support for nds32 target.
4204 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4205 * config/nds32/nds32.opt: Add -mcpu= option.
4207 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4209 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4210 isel=yes): Warn for these deprecated options.
4212 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4214 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4215 ISA_2_5_MASKS_EMBEDDED.
4217 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4219 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4220 p->max as pointers rather than using iterative_hash_expr.
4222 2018-02-23 Carl Love <cel@us.ibm.com>
4224 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4225 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4227 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4228 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4229 P8V_BUILTIN_VEC_VUNSIGNED2.
4231 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4234 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4235 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4237 * lra-constraints.c (curr_insn_transform): Set up
4238 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4240 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4241 * lra-eliminations.c (spill_pseudos): Ditto.
4242 (process_insn_for_elimination): Ditto.
4243 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4244 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4245 LRA_NON_CLOBBERED_ALT.
4247 2018-02-22 Martin Sebor <msebor@redhat.com>
4249 PR tree-optimization/84480
4250 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4251 to maybe_diag_stxncpy_trunc. Call it.
4252 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4253 from gimple_fold_builtin_strcpy. Print inlining stack.
4254 (handle_builtin_stxncpy): Print inlining stack.
4255 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4257 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4260 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4261 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4262 and -fcheck-pointer-bounds are used together.
4263 (indirect_thunk_prefix): New enum.
4264 (indirect_thunk_need_prefix): New function.
4265 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4266 "_nt" instead of "_bnd" for NOTRACK prefix.
4267 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4268 (output_indirect_thunk_function): Likewise.
4270 (ix86_code_end): Update output_indirect_thunk_function calls.
4271 (ix86_output_indirect_branch_via_reg): Replace
4272 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4273 (ix86_output_indirect_branch_via_push): Likewise.
4274 (ix86_output_function_return): Likewise.
4275 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4276 incompatible with -fcf-protection=branch and
4277 -fcheck-pointer-bounds.
4279 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4282 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4283 Change gcc_assert call to output_operand_lossage.
4285 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4287 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4289 2018-02-22 DJ Delorie <dj@redhat.com>
4290 Sebastian Perta <sebastian.perta@renesas.com>
4291 Oleg Endo <olegendo@gcc.gnu.org>
4293 * config/rx/rx.c (rx_rtx_costs): New function.
4294 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4296 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4298 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4300 2018-02-22 Martin Liska <mliska@suse.cz>
4303 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4304 Add "native" as a possible value.
4306 2018-02-22 Martin Liska <mliska@suse.cz>
4309 * config/i386/i386.c (ix86_option_override_internal):
4310 Add "native" as a possible value for -march and -mtune.
4312 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4315 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4316 to all type variants.
4318 PR tree-optimization/84503
4319 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4320 width as info->bitpos + info->bitsize - start.
4321 (merged_store_group::merge_overlapping): Simplify width computation.
4322 (check_no_overlap): New function.
4323 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4324 start + width and last_order of the group, fail if check_no_overlap
4326 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4327 to group if check_no_overlap fails.
4329 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4331 * config/rs6000/altivec.md: Delete contraint arguments to
4332 define_expand, define_split, and define_peephole2, and in
4333 define_insn_and_split if always unused.
4334 * config/rs6000/darwin.md: Ditto.
4335 * config/rs6000/dfp.md: Ditto.
4336 * config/rs6000/rs6000.md: Ditto.
4337 * config/rs6000/sync.md: Ditto.
4338 * config/rs6000/vector.md: Ditto.
4339 * config/rs6000/vsx.md: Ditto.
4341 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4343 * config/rs6000/altivec.md: Write output control strings as braced
4344 blocks instead of double-quoted strings.
4345 * config/rs6000/darwin.md: Ditto.
4346 * config/rs6000/rs6000.md: Ditto.
4347 * config/rs6000/vector.md: Ditto.
4348 * config/rs6000/vsx.md: Ditto.
4350 2018-02-21 Jason Merrill <jason@redhat.com>
4352 PR c++/84314 - ICE with templates and fastcall attribute.
4353 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4355 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4357 * ipa-cp.c (determine_versionability): Fix comment typos.
4359 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4362 * ipa-cp.c (determine_versionability): Do not version functions caling
4365 2018-02-21 Martin Liska <mliska@suse.cz>
4368 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4369 Add "native" as a possible value.
4370 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4371 the macro when native cpu detection is available.
4373 2018-02-21 Martin Liska <mliska@suse.cz>
4376 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4377 Add "native" as a possible value.
4378 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4379 when native cpu detection is available.
4381 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4382 Martin Sebor <msebor@redhat.com>
4384 PR tree-optimization/84478
4385 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4387 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4388 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4389 support which is conservatively correct, for 2 only stay conservative
4390 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4391 argument to the 2 argument get_range_strlen, adjust 6 arg
4392 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4394 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4395 (gimple_fold_builtin_strlen): Pass true as last argument to
4398 2018-02-20 Martin Sebor <msebor@redhat.com>
4401 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4402 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4403 (builtin_memref::builtin_memref): Factor out parts into
4404 set_base_and_offset and call it.
4406 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4409 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4410 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4411 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4412 search at the associated MODE_INT.
4414 2018-02-20 Jeff Law <law@redhat.com>
4417 PR tree-optimization/81592
4419 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4420 for range data rather than using global data.
4421 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4422 range data rather than using global data.
4423 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4424 pass it to children as needed.
4425 (struct directive::fmtresult): Similarly.
4426 (struct directive::set_width): Similarly.
4427 (struct directive::set_precision): Similarly.
4428 (format_integer, format_directive, parse_directive): Similarly.
4429 (format_none): Accept unnamed vr_values parameter.
4430 (format_percent, format_floating, format_character): Similarly.
4431 (format_string, format_plain): Similarly.
4432 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4433 the EVRP range analyzer for range data rather than using global data.
4434 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4435 gimple-ssa-evrp-analyze.h
4436 (class sprintf_dom_walker): Add after_dom_children member function.
4437 Add evrp_range_analyzer member.
4438 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4439 range analyzer as needed.
4440 (sprintf_dom_walker::after_dom_children): New member function.
4441 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4443 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4444 (evrp_range_analyzer::pop_to_marker): Likewise.
4446 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4448 PR tree-optimization/84419
4449 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4450 with the required type if its current type is compatible but
4453 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4456 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4457 after vectorization.
4459 2018-02-20 Martin Liska <mliska@suse.cz>
4462 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4463 possible values if we don't have a hint.
4465 2018-02-20 Martin Liska <mliska@suse.cz>
4469 * final.c (shorten_branches): Build align_tab array with one
4471 * opts.c (finish_options): Add alignment option limit check.
4472 (MAX_CODE_ALIGN): Likewise.
4473 (MAX_CODE_ALIGN_VALUE): Likewise.
4474 * doc/invoke.texi: Document maximum allowed option value for
4475 all -falign-* options.
4477 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4480 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4481 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4482 * var-tracking.c (emit_note_insn_var_location): Remove all references
4483 to NOTE_INSN_CALL_ARG_LOCATION.
4484 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4485 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4486 Use copy_rtx_if_shared.
4487 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4488 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4489 (dwarf2out_var_location): Remove handling of
4490 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4492 * final.c (final_scan_insn): Remove all references to
4493 NOTE_INSN_CALL_ARG_LOCATION.
4494 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4495 before dumping final insns.
4496 * except.c (emit_note_eh_region_end): Remove all references to
4497 NOTE_INSN_CALL_ARG_LOCATION.
4498 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4499 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4500 * config/arc/arc.c (hwloop_optimize): Likewise.
4501 * config/arm/arm.c (create_fix_barrier): Likewise.
4502 * config/s390/s390.c (s390_chunkify_start): Likewise.
4503 * config/sh/sh.c (find_barrier): Likewise.
4504 * config/i386/i386.c (rest_of_insert_endbranch,
4505 ix86_seh_fixup_eh_fallthru): Likewise.
4506 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4507 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4508 * config/frv/frv.c (frv_function_prologue): Likewise.
4509 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4511 (note_outside_basic_block_p): Remove all references to
4512 NOTE_INSN_CALL_ARG_LOCATION.
4513 * gengtype.c (adjust_field_rtx_def): Likewise.
4514 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4516 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4517 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4520 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4523 PR tree-optimization/84452
4524 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4525 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4528 2018-02-19 Martin Liska <mliska@suse.cz>
4531 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4533 2018-02-19 Martin Liska <mliska@suse.cz>
4534 Richard Sandiford <richard.sandiford@linaro.org>
4536 PR tree-optimization/82491
4537 * gimple-fold.c (get_base_constructor): Make earlier bail out
4540 2018-02-19 Carl Love <cel@us.ibm.com>
4542 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4543 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4545 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4546 P8V_BUILTIN_VEC_NEG.
4548 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4550 * config/rl78/rl78.md (movdf): New define expand.
4552 2018-02-19 Martin Liska <mliska@suse.cz>
4555 * doc/invoke.texi: Fix typo.
4556 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4558 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4560 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4561 handle rs6000_single_float and rs6000_double_float specially for
4564 2018-02-16 Jeff Law <law@redhat.com>
4566 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4569 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4572 * ipa-inline.c (inline_small_functions): Fix a typo.
4574 2018-02-16 Nathan Sidwell <nathan@acm.org>
4576 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4578 2018-02-16 Carl Love <cel@us.ibm.com>
4580 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4581 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4582 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4583 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4584 expansion to P8V_BUILTIN_VEC_FLOAT2.
4586 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4588 PR rtl-optimization/70023
4589 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4590 src_regno into account.
4592 2018-02-16 Carl Love <cel@us.ibm.com>
4594 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4595 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4596 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4597 * config/rs6000/rs6000.c: Remove case statements for
4598 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4599 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4600 and P9V_BUILTIN_VEC_VINSERT4B.
4601 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4602 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4603 * config/rs6000/vsx.md:
4604 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4607 2018-02-16 Carl Love <cel@us.ibm.com>
4609 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4611 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4613 * config/rs6000/rs6000-c.c: Add the definitions for
4614 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4615 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4616 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4617 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4618 definition for insert4b and define insn *insert3b_internal.
4619 * doc/extend.texi: Add documentation for vec_extract4b.
4621 2018-02-16 Nathan Sidwell <nathan@acm.org>
4623 * doc/extend.texi (Backwards Compatibility): Mention friend
4624 injection. Note for-scope is deprecated.
4625 * doc/invoke.texi (-ffriend-injection): Deprecate.
4627 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4629 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4630 that moved to I2, also allow destinations that are a paradoxical
4631 subreg (instead of a normal reg).
4633 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4636 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4639 2018-02-16 Richard Biener <rguenther@suse.de>
4641 PR tree-optimization/84037
4642 PR tree-optimization/84016
4644 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4645 Adjust vec_construct for the fact we need additional higher latency
4646 128bit inserts for AVX256 and AVX512 vector builds.
4647 (ix86_add_stmt_cost): Scale vector construction cost for
4650 2018-02-16 Richard Biener <rguenther@suse.de>
4652 PR tree-optimization/84417
4653 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4654 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4655 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4657 2018-02-16 Martin Liska <mliska@suse.cz>
4660 * internal-fn.def (ASAN_CHECK): Set proper flags.
4661 (ASAN_MARK): Likewise.
4663 2018-02-16 Julia Koval <julia.koval@intel.com>
4665 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4666 from PTA_CANNONLAKE.
4668 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4671 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4672 Use ++iter rather than iter++ for std::list iterators.
4673 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4674 defer deleting them until all nodes in the forest are processed. Do
4675 free even leaf nodes. Change to_process into auto_vec.
4678 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4679 * vec.h (vec_default_construct): Use memset instead of placement new
4680 if BROKEN_VALUE_INITIALIZATION is defined.
4681 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4682 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4685 PR rtl-optimization/83723
4686 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4687 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4688 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4690 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4692 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4694 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4696 PR rtl-optimization/81443
4697 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4698 from inner REGs to paradoxical SUBREGs.
4700 2018-02-16 Richard Biener <rguenther@suse.de>
4702 PR tree-optimization/84399
4703 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4704 For operands we can analyze at their definition make sure we can
4705 analyze them at each use as well.
4707 2018-02-16 Richard Biener <rguenther@suse.de>
4709 PR tree-optimization/84190
4710 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4711 volatile accesses if the decl isn't volatile.
4713 2018-02-15 Jason Merrill <jason@redhat.com>
4715 PR c++/84314 - ICE with templates and fastcall attribute.
4716 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4717 TYPE_CANONICAL on an existing type.
4719 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4721 PR tree-optimization/84383
4722 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4723 dstoff nor call operand_equal_p if dstbase is NULL.
4725 PR tree-optimization/84334
4726 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4727 also a CONSTANT_CLASS_P, punt.
4729 2018-02-14 Jim Wilson <jimw@sifive.com>
4731 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4732 first SMALL_OPERAND check. New local min_second_step. Move assert
4733 to where locals are set. Add TARGET_RVC support.
4734 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4736 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4738 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4740 2018-02-14 Martin Sebor <msebor@redhat.com>
4742 PR tree-optimization/83698
4743 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4744 arrays constrain the offset range to their bounds.
4745 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4746 (builtin_access::overlap): Avoid setting the size of overlap if it's
4748 (maybe_diag_overlap): Also consider arrays when deciding what values
4749 of offsets to include in diagnostics.
4751 2018-02-14 Martin Sebor <msebor@redhat.com>
4754 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4755 that correspond to the kind of a declaration.
4757 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4760 * config/pa/pa.md: Load address of PIC label using the linkage table
4761 if the label is nonlocal.
4763 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4765 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4766 warning message if user requests -maltivec=be.
4767 * doc/invoke.texi: Document deprecation of -maltivec=be.
4769 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4772 * config/rs6000/rs6000-c.c: Update definitions for
4773 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4774 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4776 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4779 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4780 add _get_ssp intrinsics. Remove argument from
4781 __builtin_ia32_rdssp[d|q].
4782 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4783 * config/i386/i386-builtin.def: Remove argument from
4784 __builtin_ia32_rdssp[d|q].
4785 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4786 ix86_expand_special_args_builtin for _rdssp[d|q].
4787 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4788 Clear register before usage.
4789 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4790 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4792 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4794 PR tree-optimization/84357
4795 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4796 operand 1 of an ARRAY_REF too.
4798 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4801 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4802 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4804 (set_of_reg): New struct.
4805 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4806 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4807 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4809 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4810 Split into bitclr, bitset, bitinvert patterns if appropriate.
4811 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4812 use rx_fuse_in_memory_bitop.
4813 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4814 to named insn, correct maximum insn length.
4816 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4819 * machmode.def: Define a complex mode for PARTIAL_INT.
4820 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4822 * doc/rtl.texi: Document CSPImode.
4823 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4825 (msp430_hard_regno_nregs_with_padding): Likewise.
4827 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4830 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4832 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4834 PR rtl-optimization/84169
4835 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4836 we generated a parallel as new i3 and we split that to new i2 and i3
4837 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4838 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4839 those to i2, not i1. Partially rewrite this scan code.
4841 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4844 * stor-layout.c (place_field): For variable length fields, adjust
4845 offset_align afterwards not just based on the field's alignment,
4846 but also on the size.
4849 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4850 of exps and logs in the use_exp2 case.
4852 2018-02-13 Jeff Law <law@redhat.com>
4854 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4857 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4860 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4864 * common.opt (gas-loc-support, gas-locview-support): New.
4865 (ginline-points, ginternal-reset-location-views): New.
4866 * doc/invoke.texi: Document them. Use @itemx where intended.
4867 (gvariable-location-views): Adjust.
4868 * target.def (reset_location_view): New.
4869 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4870 (TARGET_RESET_LOCATION_VIEW): New.
4871 * doc/tm.texi: Rebuilt.
4872 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4873 (dwarf2out_default_as_locview_support): New.
4874 (output_asm_line_debug_info): Use option variables.
4875 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4876 (output_loc_list): Likewise.
4877 (add_high_low_attributes): Check option variables.
4878 Don't output entry view attribute in strict mode.
4879 (gen_inlined_subroutine_die): Check option variables.
4880 (dwarf2out_inline_entry): Likewise.
4881 (init_sections_and_labels): Likewise.
4882 (dwarf2out_early_finish): Likewise.
4883 (maybe_reset_location_view): New, from...
4884 (dwarf2out_var_location): ... here. Call it.
4885 * debug.h (dwarf2out_default_as_loc_support): Declare.
4886 (dwarf2out_default_as_locview_support): Declare.
4887 * hooks.c (hook_int_rtx_insn_0): New.
4888 * hooks.h (hook_int_rtx_insn_0): Declare.
4889 * toplev.c (process_options): Take -gas-loc-support and
4890 -gas-locview-support from dwarf2out. Enable
4891 -gvariable-location-views by default only with locview
4892 assembler support. Enable -ginternal-reset-location-views by
4893 default only if the target defines the corresponding hook.
4894 Enable -ginline-points by default if location views are
4895 enabled; force it disabled if statement frontiers are
4897 * tree-inline.c (expand_call_inline): Check option variables.
4898 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4900 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4902 PR tree-optimization/84321
4903 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4904 handling. Also check whether the anti-range contains any values
4905 that satisfy the mask; switch to a VR_RANGE if not.
4907 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4910 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4912 2018-02-13 Martin Jambor <mjambor@suse.cz>
4915 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4916 of call statements, also set location of a load to a temporary.
4918 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4920 * config/rl78/rl78.c (add_vector_labels): New function.
4921 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4922 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4923 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4924 which checks that no arguments are passed.
4925 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4926 * doc/extend.texi: Documentation for the new attribute.
4928 2018-02-13 Andreas Schwab <schwab@suse.de>
4930 * config/riscv/linux.h (CPP_SPEC): Define.
4932 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4935 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4936 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4937 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4938 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4939 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4940 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4941 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4942 temporarily for AES and PCLMUL builtins.
4944 PR tree-optimization/84339
4945 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4946 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4950 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4951 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4952 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4954 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4956 * omp-simd-clone.h: New file.
4957 * omp-simd-clone.c: Include omp-simd-clone.h.
4958 (expand_simd_clones): No longer static.
4959 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4960 cgraph.h and omp-simd-clone.h.
4961 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4962 (vect_recog_widen_shift_pattern): Formatting fix.
4963 (vect_pattern_recog_1): Don't check optab for calls.
4966 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4967 operands[2] into a REG before using gen_lowpart on it.
4969 2018-02-12 Jeff Law <law@redhat.com>
4972 * config/sh/sh.c (find_barrier): Consider a sibling call
4975 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4976 successfully back substituting a reg.
4978 2018-02-12 Richard Biener <rguenther@suse.de>
4980 PR tree-optimization/84037
4981 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4982 parameter, move visited init to caller.
4983 (vect_slp_analyze_operations): Separate cost from validity
4984 check, initialize visited once for all instances.
4985 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4987 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4989 (vect_model_promotion_demotion_cost): Likewise.
4990 (vectorizable_bswap): Guard cost modeling with !slp_node
4991 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4993 (vectorizable_call): Likewise.
4994 (vectorizable_conversion): Likewise.
4995 (vectorizable_assignment): Likewise.
4996 (vectorizable_shift): Likewise.
4997 (vectorizable_operation): Likewise.
4998 (vectorizable_store): Likewise.
4999 (vectorizable_load): Likewise.
5000 (vectorizable_condition): Likewise.
5001 (vectorizable_comparison): Likewise.
5003 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5006 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5007 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5008 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5010 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5013 * params.def (inline-min-speedup): Increase from 8 to 15.
5014 (max-inline-insns-auto): Decrease from 40 to 30.
5015 * ipa-split.c (consider_split): Add some buffer for function to
5016 be considered inlining candidate.
5017 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5020 2018-02-12 Richard Biener <rguenther@suse.de>
5022 PR tree-optimization/84037
5023 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5024 matched stmts if we cannot swap the non-matched ones.
5026 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5028 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5029 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5030 _mm_maskz_scalef_round_ss): New intrinsics.
5031 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5032 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5033 __builtin_ia32_scalefss_round): Remove.
5034 (__builtin_ia32_scalefsd_mask_round,
5035 __builtin_ia32_scalefss_mask_round): New intrinsics.
5036 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5037 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5038 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5039 "<round_constraint>")): Changed to ...
5040 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5041 "<round_scalar_constraint>")): ... this.
5042 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5043 %0, %1, %2<round_op3>}"): Changed to ...
5044 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5045 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5046 %2<round_scalar_mask_op3>}"): ... this.
5047 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5049 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5051 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5052 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5053 (_mm_maskz_sqrt_round_ss): New intrinsics.
5054 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5055 (__builtin_ia32_sqrtsd_mask_round)
5056 (__builtin_ia32_sqrtss_mask_round): New builtins.
5057 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5058 (__builtin_ia32_sqrtss_round): Remove.
5059 (__builtin_ia32_sqrtsd_mask_round)
5060 (__builtin_ia32_sqrtss_mask_round): New builtins.
5061 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5062 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5063 ((match_operand:VF_128 1 "vector_operand"
5064 "xBm,<round_constraint>")): Changed to ...
5065 ((match_operand:VF_128 1 "vector_operand"
5066 "xBm,<round_scalar_constraint>")): ... this.
5067 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5068 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5069 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5070 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5071 %<iptr>1<round_scalar_mask_op3>}): ... this.
5072 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5073 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5075 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5078 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5079 Cast vec_cmpeq result to correct type.
5080 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5081 Cast vec_cmpgt result to correct type.
5083 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5085 * final.c (final_scan_insn_1): Renamed from...
5086 (final_scan_insn): ... this. New wrapper, to recover
5087 seen from the outermost call in recursive ones.
5088 * config/sparc/sparc.c (output_return): Drop seen from call.
5089 (output_sibcall): Likewise.
5090 * config/visium/visium.c (output_branch): Likewise.
5092 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5094 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5097 2018-02-10 Alan Modra <amodra@gmail.com>
5100 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5101 Specify LR as an input.
5103 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5106 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5107 remove_member_access_dummy_vars): New functions.
5108 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5109 lower_omp_1, execute_lower_omp): Use them.
5111 PR rtl-optimization/84308
5112 * shrink-wrap.c (spread_components): Release todo vector.
5114 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5116 PR rtl-optimization/57193
5117 * ira-color.c (struct allocno_color_data): Add member
5118 conflict_allocno_hard_prefs.
5119 (update_conflict_allocno_hard_prefs): New.
5120 (bucket_allocno_compare_func): Add a preference based on
5121 conflict_allocno_hard_prefs.
5122 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5123 (color_allocnos): Remove a dead code. Initiate
5124 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5126 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5129 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5130 constraint from =wa to wa. Avoid a subreg on the output operand,
5131 instead use a pseudo and subreg it in a move.
5132 (p9_xxbrd_<mode>): Changed to ...
5133 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5134 (p9_xxbrd_v2df): New expander.
5135 (p9_xxbrw_<mode>): Changed to ...
5136 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5137 (p9_xxbrw_v4sf): New expander.
5139 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5141 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5143 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5146 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5147 multiply in 32-bit mode.
5148 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5149 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5152 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5154 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5155 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5156 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5157 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5159 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5162 * dwarf2out.c (is_trivial_indirect_ref): New function.
5163 (dwarf2out_late_global_decl): Do not generate a location
5164 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5165 and that are not defined in the current unit.
5167 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5169 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5170 instead of a libcall for UNORDERED.
5172 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5175 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5176 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5178 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5181 * config/s390/s390.c (s390_set_current_function): Invoke
5182 s390_indirect_branch_settings also if fndecl didn't change.
5184 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5186 * config/rs6000/rs6000.md (blockage): Set length to zero.
5188 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5190 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5192 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5195 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5196 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5200 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5201 PARALLEL incoming that failed vt_get_decl_and_offset check.
5204 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5205 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5207 (get_variable_section): For decls in named .bss* sections pass true as
5208 second argument to bss_initializer_p.
5210 2018-02-09 Marek Polacek <polacek@redhat.com>
5211 Jakub Jelinek <jakub@redhat.com>
5214 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5215 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5216 Sync some changes from cxx_fold_indirect_ref.
5218 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5220 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5222 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5223 (BLOCK_INLINE_ENTRY_LABEL): New.
5224 (dwarf2out_var_location): Disregard inline entry markers.
5225 (inline_entry_data): New struct.
5226 (inline_entry_data_hasher): New hashtable type.
5227 (inline_entry_data_hasher::hash): New.
5228 (inline_entry_data_hasher::equal): New.
5229 (inline_entry_data_table): New variable.
5230 (add_high_low_attributes): Add DW_AT_entry_pc and
5231 DW_AT_GNU_entry_view attributes if a pending entry is found
5232 in inline_entry_data_table. Add old entry_pc attribute only
5233 if debug nonbinding markers are disabled.
5234 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5235 markers are enabled.
5236 (block_within_block_p, dwarf2out_inline_entry): New.
5237 (dwarf2out_finish): Check that no entries remained in
5238 inline_entry_data_table.
5239 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5240 (final_scan_insn, notice_source_line): Likewise.
5241 (rest_of_clean_state): Skip inline entry markers.
5242 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5244 * gimple.c (gimple_build_debug_inline_entry): New.
5245 * gimple.h (enum gimple_debug_subcode): Add
5246 GIMPLE_DEBUG_INLINE_ENTRY.
5247 (gimple_build_debug_inline_entry): Declare.
5248 (gimple_debug_inline_entry_p): New.
5249 (gimple_debug_nonbind_marker_p): Adjust.
5250 * insn-notes.def (INLINE_ENTRY): New.
5251 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5252 inline entry marker notes.
5253 (print_insn): Likewise.
5254 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5255 (INSN_DEBUG_MARKER_KIND): Likewise.
5256 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5257 * tree-inline.c (expand_call_inline): Build and insert
5258 debug_inline_entry stmt.
5259 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5260 inline entry blocks early, if nonbind markers are enabled.
5261 (dump_scope_block): Dump fragment info.
5262 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5263 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5264 (gimple_build_debug_inline_entry): New.
5265 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5266 Enable/disable inline entry points too.
5267 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5268 (DEBUG_INSN): Describe inline entry markers.
5270 * common.opt (gvariable-location-views): New.
5271 (gvariable-location-views=incompat5): New.
5272 * config.in: Rebuilt.
5273 * configure: Rebuilt.
5274 * configure.ac: Test assembler for view support.
5275 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5276 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5277 * dwarf2out.c (var_loc_view): New typedef.
5278 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5279 (dwarf2out_locviews_in_attribute): New.
5280 (dwarf2out_locviews_in_loclist): New.
5281 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5282 (enum dw_line_info_opcode): Add LI_adv_address.
5283 (struct dw_line_info_table): Add view.
5284 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5285 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5286 (zero_view_p): New variable.
5287 (ZERO_VIEW_P): New macro.
5288 (output_asm_line_debug_info): New.
5289 (struct var_loc_node): Add view.
5290 (add_AT_view_list, AT_loc_list): New.
5291 (add_var_loc_to_decl): Add view param. Test it against last.
5292 (new_loc_list): Add view params. Record them.
5293 (AT_loc_list_ptr): Handle loc and view lists.
5294 (view_list_to_loc_list_val_node): New.
5295 (print_dw_val): Handle dw_val_class_view_list.
5296 (size_of_die): Likewise.
5297 (value_format): Likewise.
5298 (loc_list_has_views): New.
5299 (gen_llsym): Set vl_symbol too.
5300 (maybe_gen_llsym, skip_loc_list_entry): New.
5301 (dwarf2out_maybe_output_loclist_view_pair): New.
5302 (output_loc_list): Output view list or entries too.
5303 (output_view_list_offset): New.
5304 (output_die): Handle dw_val_class_view_list.
5305 (output_dwarf_version): New.
5306 (output_compilation_unit_header): Use it.
5307 (output_skeleton_debug_sections): Likewise.
5308 (output_rnglists, output_line_info): Likewise.
5309 (output_pubnames, output_aranges): Update version comments.
5310 (output_one_line_info_table): Output view numbers in asm comments.
5311 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5312 Call maybe_gen_llsym.
5313 (loc_list_from_tree_1): Adjust.
5314 (add_AT_location_description): Create view list attribute if
5315 needed, check it's absent otherwise.
5316 (convert_cfa_to_fb_loc_list): Adjust.
5317 (maybe_emit_file): Call output_asm_line_debug_info for test.
5318 (dwarf2out_var_location): Reset views as needed. Precompute
5319 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5320 attribute. Set view.
5321 (new_line_info_table): Reset next view.
5322 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5323 (dwarf2out_source_line): Likewise. Output view resets and labels to
5324 the assembler, or select appropriate line info opcodes.
5325 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5326 (optimize_string_length): Catch it. Adjust.
5327 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5328 dw_val_class_view_list, and remove it if no longer needed.
5329 (hash_loc_list): Hash view numbers.
5330 (loc_list_hasher::equal): Compare them.
5331 (optimize_location_lists): Check whether a view list symbol is
5332 needed, and whether the locview attribute is present, and
5333 whether they match. Remove the locview attribute if no longer
5335 (index_location_lists): Call skip_loc_list_entry for test.
5336 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5337 Use output_dwarf_version.
5338 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5339 (struct dw_val_node): Add val_view_list.
5340 * final.c (SEEN_NEXT_VIEW): New.
5341 (set_next_view_needed): New.
5342 (clear_next_view_needed): New.
5343 (maybe_output_next_view): New.
5344 (final_start_function): Rename to...
5345 (final_start_function_1): ... this. Take pointer to FIRST,
5346 add SEEN parameter. Emit param bindings in the initial view.
5347 (final_start_function): Reintroduce SEEN-less interface.
5348 (final): Rename to...
5349 (final_1): ... this. Take SEEN parameter. Output final pending
5350 next view at the end.
5351 (final): Reintroduce seen-less interface.
5352 (final_scan_insn): Output pending next view before switching
5353 sections or ending a block. Mark the next view as needed when
5354 outputting variable locations. Notify debug backend of section
5355 changes, and of location view changes.
5356 (rest_of_handle_final): Adjust.
5357 * toplev.c (process_options): Autodetect value for debug variable
5358 location views option. Warn on incompat5 without -gdwarf-5.
5359 * doc/invoke.texi (gvariable-location-views): New.
5360 (gvariable-location-views=incompat5): New.
5361 (gno-variable-location-views): New.
5363 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5365 PR tree-optimization/84136
5366 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5367 that the result of find_edge is non-NULL.
5369 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5372 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5373 storing integer register in SImode. Fix cost of 256 and 512
5374 byte aligned SSE register store.
5376 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5378 * config/i386/i386.c (ix86_multiplication_cost): Fix
5379 multiplication cost for TARGET_AVX512DQ.
5381 2018-02-08 Marek Polacek <polacek@redhat.com>
5383 PR tree-optimization/84238
5384 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5387 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5389 PR tree-optimization/84265
5390 * tree-vect-stmts.c (vectorizable_store): Don't treat
5391 VMAT_CONTIGUOUS accesses as grouped.
5392 (vectorizable_load): Likewise.
5394 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5396 PR tree-optimization/81635
5397 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5398 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5399 (test_round_for_mask): New functions.
5400 (wide_int_cc_tests): Call test_round_for_mask.
5401 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5402 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5403 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5404 range returned by get_range_info.
5406 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5409 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5410 * symtab.c: Include builtins.h
5411 (symtab_node::output_to_lto_symbol_table_p): Move here
5412 from lto-streamer-out.c:output_symbol_p.
5413 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5414 (output_symbol_p): Move all logic to symtab.c
5415 (produce_symtab): Update.
5417 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5419 * config/s390/s390-opts.h (enum indirect_branch): Define.
5420 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5421 (s390_indirect_branch_via_thunk)
5422 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5423 (enum s390_indirect_branch_type): Define.
5424 * config/s390/s390.c (struct s390_frame_layout, struct
5425 machine_function): Remove.
5426 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5427 (indirect_branch_table_label_no, indirect_branch_table_name):
5429 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5430 (enum s390_indirect_branch_option): Define.
5431 (s390_return_addr_from_memory): New function.
5432 (s390_handle_string_attribute): New function.
5433 (s390_attribute_table): Add new attribute handler.
5434 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5435 (s390_indirect_branch_via_thunk): New function.
5436 (s390_indirect_branch_via_inline_thunk): New function.
5437 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5438 sibling call optimization for non z10 compiles.
5439 (s390_emit_call): Force indirect branch target to be a single
5440 register. Add r1 clobber for non-z10 compiles.
5441 (s390_emit_epilogue): Emit return jump via return_use expander.
5442 (s390_reorg): Handle JUMP_INSNs as execute targets.
5443 (s390_option_override_internal): Perform validity checks for the
5444 new command line options.
5445 (s390_indirect_branch_attrvalue): New function.
5446 (s390_indirect_branch_settings): New function.
5447 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5448 (s390_output_indirect_thunk_function): New function.
5449 (s390_code_end): Implement target hook.
5450 (s390_case_values_threshold): Implement target hook.
5451 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5453 * config/s390/s390.h (struct s390_frame_layout)
5454 (struct machine_function): Move here from s390.c.
5455 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5456 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5457 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5458 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5459 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5460 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5461 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5462 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5463 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5464 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5465 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5466 (mnemonic attribute): Add values which aren't recognized
5468 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5469 pattern for branch conversion. Fix mnemonic attribute.
5470 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5471 indirect branch via thunk if requested.
5472 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5473 ("*indirect_jump"): Disable for branch conversion using out of
5475 ("indirect_jump_via_thunk<mode>_z10")
5476 ("indirect_jump_via_thunk<mode>")
5477 ("indirect_jump_via_inlinethunk<mode>_z10")
5478 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5479 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5480 ("casesi_jump_via_inlinethunk<mode>_z10")
5481 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5482 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5483 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5484 ("*indirect2_jump"): Disable for branch conversion.
5485 ("casesi_jump"): Turn into expander and expand patterns for branch
5487 ("return_use"): New expander.
5488 ("*return"): Emit return via thunk and rename it to ...
5489 ("*return<mode>"): ... this one.
5490 * config/s390/s390.opt: Add new options and and enum for the
5493 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5495 * lra-constraints.c (match_reload): Unconditionally use
5496 gen_lowpart_SUBREG, rather than selecting between that
5497 and equivalent gen_rtx_SUBREG code.
5499 2018-02-08 Richard Biener <rguenther@suse.de>
5501 PR tree-optimization/84233
5502 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5503 changed flag instead of boguously re-using phi_inserted.
5505 2018-02-08 Martin Jambor <mjambor@suse.cz>
5507 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5508 static local variables.
5510 2018-02-08 Richard Biener <rguenther@suse.de>
5512 PR tree-optimization/84278
5513 * tree-vect-stmts.c (vectorizable_store): When looking for
5514 smaller vector types to perform grouped strided loads/stores
5515 make sure the mode is supported by the target.
5516 (vectorizable_load): Likewise.
5518 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5520 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5521 Increase LDP/STP opportunities by adding adjacent callee-saves.
5523 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5525 PR rtl-optimization/84068
5526 PR rtl-optimization/83459
5527 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5529 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5531 PR tree-optimization/84224
5532 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5533 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5536 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5539 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5540 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5541 count, remove one USE.
5543 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5545 * doc/install.texi (Configuration): Document the
5546 --with-long-double-format={ibm,ieee} PowerPC configuration
5550 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5551 Convert from define_expand to be define_insn_and_split. Rework
5552 float/double/_Float128 conversions to QI/HI/SImode to work with
5553 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5554 conversions to QI/HImode types did a store and then a load to
5555 truncate the value. For conversions to VSX registers, don't split
5556 the insn, instead emit the code directly. Use the code iterator
5557 any_fix to combine signed and unsigned conversions.
5558 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5559 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5560 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5561 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5562 (fix_<mode>di2_hw): Likewise.
5563 (fixuns_<mode>di2_hw): Likewise.
5564 (fix_<mode>si2_hw): Likewise.
5565 (fixuns_<mode>si2_hw): Likewise.
5566 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5567 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5568 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5569 fix<uns>_trunc<SFDF:mode>si2_p8.
5570 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5572 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5573 (fix<uns>_<mode>_mem): Likewise.
5574 (fctiw<u>z_<mode>_mem): Likewise.
5575 (fix<uns>_<mode>_mem): Likewise.
5576 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5577 the register allocator from doing a direct move to the GPRs to do
5578 a store, and instead use the ISA 3.0 store byte/half-word from
5579 vector register instruction. For IEEE 128-bit floating point,
5580 also optimize stores of 32-bit ints.
5581 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5583 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5585 * genextract.c (push_pathstr_operand): New function to support
5587 (walk_rtx): Call push_pathstr_operand.
5588 (print_path): Support [a-zA-Z].
5590 2018-02-07 Richard Biener <rguenther@suse.de>
5592 PR tree-optimization/84037
5593 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5594 (cse_and_gimplify_to_preheader): Declare.
5595 (vect_get_place_in_interleaving_chain): Likewise.
5596 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5598 (_loop_vec_info::~_loop_vec_info): Delete it.
5599 (cse_and_gimplify_to_preheader): New function.
5600 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5601 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5602 (vectorizable_load): Likewise. For grouped stores always base
5603 the IV on the first element.
5604 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5605 condition before gimplifying.
5607 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5609 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5610 *DIV_EXPR and *MOD_EXPR.
5612 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5615 * config/i386/i386.c (ix86_option_override_internal): Mask out
5616 the CF_SET bit when checking -fcf-protection.
5618 2018-02-07 Tom de Vries <tom@codesourcery.com>
5621 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5624 2018-02-07 Richard Biener <rguenther@suse.de>
5626 PR tree-optimization/84204
5627 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5630 PR tree-optimization/84205
5631 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5632 special-case isl_ast_op_zdiv_r.
5634 PR tree-optimization/84223
5635 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5636 Only add conditions from within the region.
5637 (gather_bbs::after_dom_children): Adjust.
5639 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5642 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5643 * config/avr/avr.md: Only post-reload split REG-REG moves if
5644 either register is GENERAL_REG_P.
5646 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5648 PR tree-optimization/84235
5649 * tree-ssa-scopedtables.c
5650 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5651 if the subtraction is performed in floating point type where NaNs are
5652 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5653 build 1. Formatting fix.
5655 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5658 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5659 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5660 and skip it regardless of bb boundaries. Use CALL_P macro,
5661 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5662 unnecessarily, formatting fix.
5664 2018-02-06 Michael Collison <michael.collison@arm.com>
5666 * config/arm/thumb2.md:
5667 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5668 (*thumb_mov_notscc): Ditto.
5670 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5673 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5674 unsigned_fix, not "s".
5676 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5678 * configure.ac (gcc_fn_eh_frame_ro): New function.
5679 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5680 correct .eh_frame permissions.
5681 * configure: Regenerate.
5683 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5685 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5688 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5690 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5691 Display warning message for -mno-speculate-indirect-jumps.
5693 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5695 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5697 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5699 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5701 PR tree-optimization/84225
5702 * tree-eh.c (find_trapping_overflow): Only call
5703 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5705 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5708 * config/i386/i386.c: Reimplement the check of possible options
5709 -mibt/-mshstk conbination. Change error messages.
5710 * doc/invoke.texi: Fix a typo: remove extra '='.
5712 2018-02-06 Marek Polacek <polacek@redhat.com>
5714 PR tree-optimization/84228
5715 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5717 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5720 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5721 emitted arch directives.
5722 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5723 __ARM_FEATURE_COPROC before changing architectures.
5725 2018-02-06 Richard Biener <rguenther@suse.de>
5727 * config/i386/i386.c (print_reg): Fix typo.
5728 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5730 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5732 * configure: Regenerate.
5734 2018-02-05 Martin Sebor <msebor@redhat.com>
5736 PR tree-optimization/83369
5737 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5740 2018-02-05 Martin Liska <mliska@suse.cz>
5742 * doc/invoke.texi: Cherry-pick upstream r323995.
5744 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5746 * ira.c (ira_init_register_move_cost): Adjust comment.
5748 2018-02-05 Martin Liska <mliska@suse.cz>
5750 PR gcov-profile/84137
5751 * doc/gcov.texi: Fix typo in documentation.
5753 2018-02-05 Martin Liska <mliska@suse.cz>
5755 PR gcov-profile/83879
5756 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5757 using dlopen functionality.
5759 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5761 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5762 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5763 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5764 _mm_maskz_range_round_ss): New intrinsics.
5765 (__builtin_ia32_rangesd128_round)
5766 (__builtin_ia32_rangess128_round): Remove.
5767 (__builtin_ia32_rangesd128_mask_round,
5768 __builtin_ia32_rangess128_mask_round): New builtins.
5769 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5770 __builtin_ia32_rangess128_round): Remove.
5771 (__builtin_ia32_rangesd128_mask_round,
5772 __builtin_ia32_rangess128_mask_round): New builtins.
5773 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5774 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5775 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5776 "<round_saeonly_constraint>")): Changed to ...
5777 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5778 "<round_saeonly_scalar_constraint>")): ... this.
5779 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5780 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5781 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5782 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5783 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5785 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5787 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5789 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5790 Remove all values except native, 8540 and 8548.
5792 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5794 * config/i386/i386.c (ix86_output_function_return): Pass
5795 INVALID_REGNUM, instead of -1, as invalid register number to
5796 indirect_thunk_name and output_indirect_thunk.
5798 2018-02-02 Julia Koval <julia.koval@intel.com>
5800 * config.gcc: Add -march=icelake.
5801 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5802 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5803 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5804 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5805 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5806 (processor_target_table): Add icelake.
5807 (ix86_option_override_internal): Handle new PTAs.
5808 (get_builtin_code_for_version): Handle icelake.
5809 (M_INTEL_COREI7_ICELAKE): New.
5810 (fold_builtin_cpu): Handle icelake.
5811 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5812 * doc/invoke.texi: Add -march=icelake.
5814 2018-02-02 Julia Koval <julia.koval@intel.com>
5816 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5817 to wide_int_bitmask.
5818 * wide-int-bitmask.h: New.
5820 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5823 * config/i386/i386.md: Replace Pmode with word_mode in
5824 builtin_setjmp_setup and builtin_longjmp to support x32.
5826 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5830 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5832 (rs6000_supported_cpu_names): New static variable.
5833 (linux_cpu_translation_table): Likewise.
5834 (elf_platform) <cpu>: Define new static variable and use it.
5835 Translate kernel AT_PLATFORM name to canonical name if needed.
5836 Error if platform name is unknown.
5838 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5841 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5843 2018-02-01 Jeff Law <law@redhat.com>
5846 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5847 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5848 the scratch if RELEASE_VIA_POP is false.
5849 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5850 If we have to save a temporary register, decrement SIZE appropriately.
5851 Pass new arguments to release_scratch_register_on_entry.
5852 (ix86_adjust_stack_and_probe): Likewise.
5853 (ix86_emit_probe_stack_range): Pass new arguments to
5854 release_scratch_register_on_entry.
5856 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5858 PR rtl-optimization/84157
5859 * combine.c (change_zero_ext): Use REG_P predicate in
5860 front of HARD_REGISTER_P predicate.
5862 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5864 * config/avr/avr.c (avr_option_override): Move disabling of
5865 -fdelete-null-pointer-checks to...
5866 * common/config/avr/avr-common.c (avr_option_optimization_table):
5869 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5871 PR tree-optimization/81635
5872 * tree-data-ref.c (split_constant_offset_1): For types that
5873 wrap on overflow, try to use range info to prove that wrapping
5876 2018-02-01 Renlin Li <renlin.li@arm.com>
5879 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5881 (aarch64_register_move_cost): Likewise.
5882 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5884 (REG_CLASS_NAMES): Likewise.
5885 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5886 TAILCALL_ADDR_REGS. Remove IP registers.
5887 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5889 2018-02-01 Richard Biener <rguenther@suse.de>
5891 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5892 for specifying RPO order and allow NULL for that.
5893 * domwalk.c (dom_walker::dom_walker): Likewise.
5894 (dom_walker::walk): Handle NULL RPO order.
5895 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5897 (rewrite_update_dom_walker): Likewise.
5898 (mark_def_dom_walker): Likewise.
5900 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5902 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5903 (aarch64_maybe_expand_sve_subreg_move): Declare.
5904 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5905 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5907 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5908 that are semantically a reverse operation.
5909 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5910 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5911 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5913 (aarch64_can_change_mode_class): For big-endian, forbid changes
5914 between two SVE modes if they have different element sizes.
5916 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5918 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5919 the TImode handling for big-endian targets.
5921 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5923 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5924 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5926 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5927 Remove BSWAP handing for big-endian targets and use the form of
5928 LD1RQ appropariate for the mode.
5930 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5932 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5933 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5936 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5939 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5940 check for operands that need to go through aarch64_sve_reload_be.
5942 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5944 PR tree-optimization/81661
5945 PR tree-optimization/84117
5946 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5947 * tree-eh.c: Include gimplify.h.
5948 (find_trapping_overflow, replace_trapping_overflow,
5949 rewrite_to_non_trapping_overflow): New functions.
5950 * tree-vect-loop.c: Include tree-eh.h.
5951 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5952 * tree-data-ref.c: Include tree-eh.h.
5953 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5955 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5957 PR rtl-optimization/84123
5958 * combine.c (change_zero_ext): Check if hard register satisfies
5959 can_change_dest_mode before calling gen_lowpart_SUBREG.
5961 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5964 * ira.c (ira_init_register_move_cost): Remove assert.
5966 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5968 PR rtl-optimization/84071
5969 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5970 * doc/tm.texi: Regenerate.
5972 2018-01-31 Richard Biener <rguenther@suse.de>
5974 PR tree-optimization/84132
5975 * tree-data-ref.c (analyze_miv_subscript): Properly
5976 check whether evolution_function_is_affine_multivariate_p
5977 before calling gcd_of_steps_may_divide_p.
5979 2018-01-31 Julia Koval <julia.koval@intel.com>
5982 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5983 * config/i386/i386.md (rdpid_rex64) New.
5984 (rdpid): Make 32bit only.
5986 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5989 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5990 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5992 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5995 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5997 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5999 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6001 PR rtl-optimization/84071
6002 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6003 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6005 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6007 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6008 (arc_attribute_table): Add 'aux' attribute.
6009 (arc_in_small_data_p): Consider aux like variables.
6010 (arc_is_aux_reg_p): New function.
6011 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6012 (arc_get_aux_arg): New function.
6013 (prepare_move_operands): Handle aux-register access.
6014 (arc_handle_aux_attribute): New function.
6015 * doc/extend.texi (ARC Variable attributes): Add subsection.
6017 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6019 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6020 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6021 (arc_attribute_table): Add 'uncached' attribute.
6022 (arc_print_operand): Print '.di' flag for uncached memory
6024 (arc_in_small_data_p): Do not consider for small data the uncached
6026 (arc_is_uncached_mem_p): New function.
6027 * config/arc/predicates.md (compact_store_memory_operand): Check
6028 for uncached memory accesses.
6029 (nonvol_nonimm_operand): Likewise.
6030 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6032 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6035 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6036 falign-loops=): Add Optimization flag.
6038 2018-01-30 Jeff Law <law@redhat.com>
6041 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6042 INT_REGISTERS_SAVED. Check it prior to calling
6043 get_scratch_register_on_entry.
6044 (ix86_adjust_stack_and_probe): Similarly.
6045 (ix86_emit_probe_stack_range): Similarly.
6046 (ix86_expand_prologue): Corresponding changes.
6048 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6051 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6052 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6054 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6057 * lra-constraints.c (curr_insn_transform): Process AND in the
6060 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6062 PR rtl-optimization/83986
6063 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6064 dependence against last_pending_memory_flush in addition to
6067 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6069 PR tree-optimization/81611
6070 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6073 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6076 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6079 2018-01-30 Richard Biener <rguenther@suse.de>
6080 Jakub Jelinek <jakub@redhat.com>
6082 PR tree-optimization/84111
6083 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6084 inner loops added during recursion, as they don't have up-to-date
6087 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6090 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6091 (can_inline_edge_by_limits_p): ... here.
6092 (can_early_inline_edge_p, check_callers,
6093 update_caller_keys, update_callee_keys, recursive_inlining,
6094 add_new_edges_to_heap, speculation_useful_p,
6095 inline_small_functions,
6096 inline_small_functions, flatten_function,
6097 inline_to_all_callers_1): Update.
6099 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6101 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6104 2018-01-30 Richard Biener <rguenther@suse.de>
6106 PR tree-optimization/83008
6107 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6108 invariant and constant vector uses in stmts when they need
6111 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6114 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6115 * configure: Regenerate.
6117 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6119 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6121 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6122 Use gen_rtx_REG rather than gen_lowpart.
6124 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6126 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6127 rather than 0 when creating partial subregs.
6129 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6131 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6134 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6137 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6138 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6139 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6140 flags. This restores the settings used before the 2017-07-24.
6141 Turning off pre increment/decrement/modify allows IVOPTS to
6142 optimize DF/SF loops where the index is an int.
6144 2018-01-29 Richard Biener <rguenther@suse.de>
6145 Kelvin Nilsen <kelvin@gcc.gnu.org>
6148 * tree-vect-stmts.c (vectorizable_call): Don't call
6149 targetm.vectorize_builtin_md_vectorized_function if callee is
6152 2018-01-22 Carl Love <cel@us.ibm.com>
6154 * doc/extend.tex: Fix typo in second arg in
6155 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6157 2018-01-29 Richard Biener <rguenther@suse.de>
6159 PR tree-optimization/84086
6160 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6161 (flush_ssaname_freelist): When SSA names were released reset
6162 the SCEV hash table.
6164 2018-01-29 Richard Biener <rguenther@suse.de>
6166 PR tree-optimization/84057
6167 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6168 removed paths when removing edges.
6170 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6172 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6173 -mfunction-return=@var{choice}.
6175 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6178 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6180 (layout::print_source_line): Likewise.
6181 (test_get_line_width_without_trailing_whitespace): Add test cases.
6183 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6186 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6189 2018-01-26 Jim Wilson <jimw@sifive.com>
6191 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6193 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6196 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6198 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6199 and CMP + SUB-immediate -> SUBS.
6201 2018-01-26 Martin Sebor <msebor@redhat.com>
6203 PR tree-optimization/83896
6204 * tree-ssa-strlen.c (get_string_len): Rename...
6205 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6206 Avoid assuming length is constant.
6207 (handle_char_store): Use HOST_WIDE_INT for string length.
6209 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6212 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6213 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6215 2018-01-26 Richard Biener <rguenther@suse.de>
6217 PR rtl-optimization/84003
6218 * dse.c (record_store): Only record redundant stores when
6219 the earlier store aliases at least all accesses the later one does.
6221 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6223 PR rtl-optimization/83985
6224 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6225 REG_CFA_RESTORE insns.
6226 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6227 REG_CFA_RESTORE insns here.
6230 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6231 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6233 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6235 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6236 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6237 (arc_init): Likewise.
6238 (arc_override_options): Likewise.
6239 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6241 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6243 * config/arc/arc.h (TARGET_DBNZ): Define.
6244 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6245 properly set the tune attribute.
6246 (dbnz): Use TARGET_DBNZ guard.
6247 * config/arc/arc.opt (mtune): Add core3 option.
6249 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6251 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6252 recognize new pic like addresses.
6253 (arc_delegitimize_address): Clean up.
6255 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6257 * config/arc/arc-arches.def: Option mrf16 valid for all
6259 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6260 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6261 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6262 * config/arc/arc-tables.opt: Regenerate.
6263 * config/arc/arc.c (arc_conditional_register_usage): Handle
6264 reduced register file case.
6265 (arc_file_start): Set must have build attributes.
6266 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6268 * config/arc/arc.opt (mrf16): Add new option.
6269 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6270 * config/arc/genmultilib.awk: Handle new mrf16 option.
6271 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6272 * config/arc/t-multilib: Regenerate.
6273 * doc/invoke.texi (ARC Options): Document mrf16 option.
6275 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6277 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6278 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6279 (arc_attribute_table): Add 'secure_call' attribute.
6280 (arc_print_operand): Print secure call operand.
6281 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6283 (arc_is_secure_call_p): New function. * config/arc/arc.md
6284 (call_i): Add support for sjli instruction.
6285 (call_value_i): Likewise.
6286 * config/arc/constraints.md (Csc): New constraint.
6288 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6289 John Eric Martin <John.Martin@emmicro-us.com>
6291 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6292 * config/arc/arc.c (_arc_jli_section): New struct.
6293 (arc_jli_section): New type.
6294 (rc_jli_sections): New static variable.
6295 (arc_handle_jli_attribute): New function.
6296 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6297 (arc_file_end): New function.
6298 (TARGET_ASM_FILE_END): Define.
6299 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6300 (arc_add_jli_section): New function.
6301 (jli_call_scan): Likewise.
6302 (arc_reorg): Call jli_call_scan.
6303 (arc_output_addsi): Remove 'S' from printing asm operand.
6304 (arc_is_jli_call_p): New function.
6305 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6307 (movhi_insn): Likewise.
6308 (movsi_insn): Likewise.
6309 (movsi_set_cc_insn): Likewise.
6310 (loadqi_update): Likewise.
6311 (load_zeroextendqisi_update): Likewise.
6312 (load_signextendqisi_update): Likewise.
6313 (loadhi_update): Likewise.
6314 (load_zeroextendhisi_update): Likewise.
6315 (load_signextendhisi_update): Likewise.
6316 (loadsi_update): Likewise.
6317 (loadsf_update): Likewise.
6318 (movsicc_insn): Likewise.
6319 (bset_insn): Likewise.
6320 (bxor_insn): Likewise.
6321 (bclr_insn): Likewise.
6322 (bmsk_insn): Likewise.
6323 (bicsi3_insn): Likewise.
6324 (cmpsi_cc_c_insn): Likewise.
6325 (movsi_ne): Likewise.
6326 (movsi_cond_exec): Likewise.
6327 (clrsbsi2): Likewise.
6338 (call_i): Remove 'S' asm letter, add jli instruction.
6339 (call_value_i): Likewise.
6340 * config/arc/arc.op (mjli-always): New option.
6341 * config/arc/constraints.md (Cji): New constraint.
6342 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6344 (subsf3_fpx): Likewise.
6345 (mulsf3_fpx): Likewise.
6346 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6348 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6350 * doc/invoke.texi (ARC): Document mjli-always option.
6352 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6354 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6355 avoid addition with 0 and use incw and decw where possible.
6357 2018-01-26 Richard Biener <rguenther@suse.de>
6359 PR tree-optimization/81082
6360 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6361 association if it requires casting to unsigned.
6362 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6363 from fold_plusminus_mult_expr to catch important cases late when
6364 range info is available.
6366 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6368 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6369 * configure.ac (hidden_linkonce): New test.
6370 * configure: Regenerate.
6371 * config.in: Regenerate.
6373 2018-01-26 Julia Koval <julia.koval@intel.com>
6375 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6376 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6377 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6378 _mm_mask_bitshuffle_epi64_mask): Fix type.
6379 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6380 USI_FTYPE_V4DI_V4DI_USI): Remove.
6381 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6382 __builtin_ia32_vpshufbitqmb256_mask,
6383 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6384 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6385 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6387 2018-01-26 Alan Modra <amodra@gmail.com>
6390 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6391 UNSPEC_VBPERMQ. Sort other unspecs.
6393 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6395 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6397 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6400 * predict.c (drop_profile): Do not push/pop cfun; update also
6402 (handle_missing_profiles): Fix logic looking for zero profiles.
6404 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6407 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6408 on functions with #pragma omp declare simd or functions with simd
6410 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6411 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6412 Remove trailing \n from warning_at calls.
6414 2018-01-25 Tom de Vries <tom@codesourcery.com>
6417 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6418 for neutered workers.
6420 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6423 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6424 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6426 2018-01-24 Jeff Law <law@redhat.com>
6429 * i386.c (get_probe_interval): Move to earlier point.
6430 (ix86_compute_frame_layout): If -fstack-clash-protection and
6431 the frame is larger than the probe interval, then use pushes
6432 to save registers rather than reg->mem moves.
6433 (ix86_expand_prologue): Remove conditional for int_registers_saved
6436 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6439 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6440 min/max for never referenced object.
6442 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6445 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6447 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6448 attributes from DECL_ATTRIBUTES (decl) without affecting
6449 DECL_ATTRIBUTES (current_function_decl).
6450 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6451 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6453 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6455 PR tree-optimization/83979
6456 * fold-const.c (fold_comparison): Use constant_boolean_node
6457 instead of boolean_{true,false}_node.
6459 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6461 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6464 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6466 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6467 Simplify the clause that sets the length attribute.
6468 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6469 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6470 clause that sets the length attribute.
6471 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6473 2018-01-24 Tom de Vries <tom@codesourcery.com>
6476 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6477 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6478 Add strict parameter.
6479 (prevent_branch_around_nothing): Insert dummy insn between branch to
6480 label and label with no ptx insn inbetween.
6481 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6483 2018-01-24 Tom de Vries <tom@codesourcery.com>
6486 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6487 for neutered threads in warp.
6488 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6490 2018-01-24 Richard Biener <rguenther@suse.de>
6492 PR tree-optimization/83176
6493 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6496 2018-01-24 Richard Biener <rguenther@suse.de>
6498 PR tree-optimization/82819
6499 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6500 code generating pluses that are no-ops in the target precision.
6502 2018-01-24 Richard Biener <rguenther@suse.de>
6505 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6507 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6509 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6510 to merge probabilities.
6511 * predict.c (probably_never_executed): Also mark as cold functions
6512 with global 0 profile and guessed local profile.
6513 * profile-count.c (profile_probability::combine_with_count): New
6515 * profile-count.h (profile_probability::operator*,
6516 profile_probability::operator*=, profile_probability::operator/,
6517 profile_probability::operator/=): Reduce precision to adjusted
6518 and set value to guessed on contradictory divisions.
6519 (profile_probability::combine_with_freq): Remove.
6520 (profile_probability::combine_wiht_count): Declare.
6521 (profile_count::force_nonzero):: Set to adjusted.
6522 (profile_count::probability_in):: Set quality to adjusted.
6523 * tree-ssa-tail-merge.c (replace_block_by): Use
6526 2018-01-23 Andrew Waterman <andrew@sifive.com>
6527 Jim Wilson <jimw@sifive.com>
6529 * config/riscv/riscv.c (riscv_stack_boundary): New.
6530 (riscv_option_override): Set riscv_stack_boundary. Handle
6531 riscv_preferred_stack_boundary_arg.
6532 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6533 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6534 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6535 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6536 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6537 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6539 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6542 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6543 of struct ix86_frame.
6544 (ix86_expand_epilogue): Likewise. Add a local variable for
6545 the reg_save_offset field in struct ix86_frame.
6547 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6549 PR tree-optimization/82604
6550 * tree-loop-distribution.c (enum partition_kind): New enum item
6551 PKIND_PARTIAL_MEMSET.
6552 (partition_builtin_p): Support above new enum item.
6553 (generate_code_for_partition): Ditto.
6554 (compute_access_range): Differentiate cases that equality can be
6555 proven at all loops, the innermost loops or no loops.
6556 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6557 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6558 (finalize_partitions, distribute_loop): Don't fuse partition of
6559 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6560 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6563 2018-01-23 Martin Liska <mliska@suse.cz>
6565 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6566 order to ignore the predictor.
6567 (PRED_POLYMORPHIC_CALL): Likewise.
6568 (PRED_RECURSIVE_CALL): Likewise.
6570 2018-01-23 Martin Liska <mliska@suse.cz>
6572 * tree-profile.c (tree_profiling): Print function header to
6573 aware reader which function we are working on.
6574 * value-prof.c (gimple_find_values_to_profile): Do not print
6575 not interesting value histograms.
6577 2018-01-23 Martin Liska <mliska@suse.cz>
6579 * profile-count.h (enum profile_quality): Add
6580 profile_uninitialized as the first value. Do not number values
6581 as they are zero based.
6582 (profile_count::verify): Update sanity check.
6583 (profile_probability::verify): Likewise.
6585 2018-01-23 Nathan Sidwell <nathan@acm.org>
6587 * doc/invoke.texi (ffor-scope): Deprecate.
6589 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6591 PR tree-optimization/83510
6592 * domwalk.c (set_all_edges_as_executable): New function.
6593 (dom_walker::dom_walker): Convert bool param
6594 "skip_unreachable_blocks" to enum reachability. Move setup of
6595 edge flags to set_all_edges_as_executable and only do it when
6596 reachability is REACHABLE_BLOCKS.
6597 * domwalk.h (enum dom_walker::reachability): New enum.
6598 (dom_walker::dom_walker): Convert bool param
6599 "skip_unreachable_blocks" to enum reachability.
6600 (set_all_edges_as_executable): New decl.
6601 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6602 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6604 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6605 but converting true to REACHABLE_BLOCKS.
6606 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6608 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6609 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6610 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6612 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6613 if check_all_array_refs will be called.
6615 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6617 * tree.c (selftest::test_location_wrappers): Add more test
6620 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6622 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6623 (selftest::test_bit_in_range): Likewise.
6625 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6628 * doc/sourcebuild.texi (vect_float): Say that the selector
6629 only describes the situation when -funsafe-math-optimizations is on.
6630 (vect_float_strict): Document.
6632 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6634 PR tree-optimization/83965
6635 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6636 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6637 instead of checking only for a reduction.
6638 (vect_recog_widen_sum_pattern): Likewise.
6640 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6642 * predict.c (probably_never_executed): Only use precise profile info.
6643 (compute_function_frequency): Skip after inlining hack since we now
6644 have quality checking.
6646 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6648 * profile-count.h (profile_probability::very_unlikely,
6649 profile_probability::unlikely, profile_probability::even): Set
6650 precision to guessed.
6652 2018-01-23 Richard Biener <rguenther@suse.de>
6654 PR tree-optimization/83963
6655 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6656 Properly terminate dominator walk when crossing the exit edge not
6657 when visiting its source block.
6659 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6662 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6663 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6665 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6667 PR tree-optimization/83957
6668 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6669 semicolon after for body surrounded by braces.
6671 PR tree-optimization/83081
6672 * profile-count.h (profile_probability::split): New method.
6673 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6674 Use profile_probability::split.
6675 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6676 when splitting a single conditional jump into 2.
6678 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6680 PR tree-optimization/69452
6681 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6684 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6686 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6687 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6688 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6690 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6692 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6693 * config/rl78/rl78.md (movdi): New define_expand.
6694 * config/rl78/rl78.c (rl78_split_movdi): New function.
6696 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6699 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6701 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6702 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6703 128-bit to produce an UNSPEC move to get the double word with the
6704 signbit and then a shift directly to do signbit.
6705 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6706 implementation with a new version that just does either a direct
6707 move or a regular move. Move memory interface to separate insns.
6708 Move insns so they are next to the expander.
6709 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6710 with signbit move. Split big and little endian case.
6711 (signbit<mode>2_dm_mem_le): Likewise.
6712 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6713 (signbit<mode>2_dm2): Likewise.
6715 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6717 * config/rl78/rl78.md (anddi3): New define_expand.
6719 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6721 * config/rl78/rl78.md (umindi3): New define_expand.
6723 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6725 * config/rl78/rl78.md (smindi3): New define_expand.
6727 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6729 * config/rl78/rl78.md (smaxdi3): New define_expand.
6731 2018-01-22 Carl Love <cel@us.ibm.com>
6733 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6734 LVX_V1TI): Add macro expansion.
6735 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6736 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6737 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6738 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6739 Change check to determine if the instruction is a byte reversing
6740 entry. Fix typo in comment.
6741 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6742 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6743 Add def_builtin calls for new builtins.
6744 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6745 Add define_insn expansion.
6747 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6749 * config/rl78/rl78.md (umaxdi3): New define_expand.
6751 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6753 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6754 for non-QImode registers.
6756 2018-01-22 Richard Biener <rguenther@suse.de>
6758 PR tree-optimization/83963
6759 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6760 including the loop exit block.
6761 (scop_detection::merge_sese): Likewise.
6762 (scop_detection::add_scop): Do it here instead.
6764 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6766 * doc/sourcebuild.texi (arm_softfloat): Document.
6768 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6771 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6772 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6773 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6775 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6776 David Edelsohn <dje.gcc@gmail.com>
6779 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6780 Change "crset eq" to "crset 2".
6781 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6782 (*call_indirect_aix<mode>_nospec): Likewise.
6783 (*call_value_indirect_aix<mode>_nospec): Likewise.
6784 (*call_indirect_elfv2<mode>_nospec): Likewise.
6785 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6786 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6787 change assembly output from . to $.
6788 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6789 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6790 (*tablejump<mode>_internal1_nospec): Likewise.
6792 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6795 * config/sh/sh_optimize_sett_clrt.cc:
6796 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6798 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6800 PR tree-optimization/83940
6801 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6802 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6803 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6804 use it to pass back the definition type.
6805 (vect_check_store_rhs): Likewise rhs_dt_out.
6806 (vect_build_gather_load_calls): Add a mask_dt argument and use
6807 it instead of a call to vect_is_simple_use.
6808 (vectorizable_store): Update calls to vect_check_load_store_mask
6809 and vect_check_store_rhs. Use the dt returned by the latter instead
6810 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6811 instead of calls to vect_is_simple_use. Pass the scalar rather
6812 than the vector operand to vect_is_simple_use when handling
6813 second and subsequent copies of an rhs value.
6814 (vectorizable_load): Update calls to vect_check_load_store_mask
6815 and vect_build_gather_load_calls. Use the cached mask_dt and
6816 gs_info.offset_dt instead of calls to vect_is_simple_use.
6818 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6821 * tree-emutls.c: Include gimplify.h.
6822 (lower_emutls_2): New function.
6823 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6824 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6825 it before further processing.
6828 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6829 UINTVAL (trueop1) instead of INTVAL (op1).
6831 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6835 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6836 INCOMING_FRAME_SP_OFFSET if not defined.
6837 (scan_trace): Add ENTRY argument. If true and
6838 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6839 emit a note to adjust the CFA offset.
6840 (create_cfi_notes): Adjust scan_trace callers.
6841 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6842 INCOMING_FRAME_SP_OFFSET in the CIE.
6843 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6844 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6846 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6847 * doc/tm.texi: Regenerated.
6849 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6851 PR rtl-optimization/83147
6852 * lra-constraints.c (remove_inheritance_pseudos): Use
6853 lra_substitute_pseudo_within_insn.
6855 2018-01-19 Tom de Vries <tom@codesourcery.com>
6856 Cesar Philippidis <cesar@codesourcery.com>
6859 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6861 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6864 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6865 spaces for function labels.
6867 2018-01-19 Martin Liska <mliska@suse.cz>
6869 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6870 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6871 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6872 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6873 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6874 (PRED_CONST_RETURN): Change from 69 to 65.
6875 (PRED_NULL_RETURN): Change from 91 to 71.
6876 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6877 (PRED_LOOP_GUARD): Change from 66 to 73.
6879 2018-01-19 Martin Liska <mliska@suse.cz>
6881 * predict.c (predict_insn_def): Add new assert.
6882 (struct branch_predictor): Change type to signed integer.
6883 (test_prediction_value_range): Amend test to cover
6885 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6886 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6887 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6888 (PRED_LOOP_IV_COMPARE): Likewise.
6889 * predict.h (PROB_UNINITIALIZED): Define new constant.
6891 2018-01-19 Martin Liska <mliska@suse.cz>
6893 * predict.c (dump_prediction): Add new format for
6894 analyze_brprob.py script which is enabled with -details
6896 * profile-count.h (precise_p): New function.
6898 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6900 PR tree-optimization/83922
6901 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6902 there are no statements that need masking.
6903 (vect_active_double_reduction_p): New function.
6904 (vect_analyze_loop_operations): Use it when handling phis that
6905 are not in the loop header.
6907 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6909 PR tree-optimization/83914
6910 * tree-vect-loop.c (vectorizable_induction): Don't convert
6911 init_expr or apply the peeling adjustment for inductions
6912 that are nested within the vectorized loop.
6914 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6916 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6919 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6923 * function.h (gimplify_parameters): Add gimple_seq * argument.
6924 * function.c: Include gimple.h and options.h.
6925 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6926 for the added local temporaries if needed.
6927 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6928 if there are any parameter cleanups, wrap whole body into a
6929 try/finally with the cleanups.
6931 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6934 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6935 Use GET_MODE_CLASS for scalar floating point.
6937 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6941 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6942 Fix call of call_cgraph_insertion_hooks.
6944 2018-01-18 Martin Sebor <msebor@redhat.com>
6946 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6948 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6951 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6954 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6957 * common.opt: (-ffile-prefix-map): New option.
6958 * opts.c (common_handle_option): Defer it.
6959 * opts-global.c (handle_common_deferred_options): Handle it.
6960 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6961 * file-prefix-map.h: New file.
6962 (remap_debug_filename, add_debug_prefix_map): ...here.
6963 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6964 * final.c (debug_prefix_map, add_debug_prefix_map
6965 remap_debug_filename): Move to...
6966 * file-prefix-map.c: New file.
6967 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6968 generalize, get rid of alloca(), use strrchr() instead of strchr().
6969 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6970 Implement in terms of add_prefix_map().
6971 (remap_macro_filename, remap_debug_filename): Implement in term of
6973 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6974 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6975 * dbxout.c: Include file-prefix-map.h.
6976 * varasm.c: Likewise.
6977 * vmsdbgout.c: Likewise.
6978 * xcoffout.c: Likewise.
6979 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6980 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6981 * doc/invoke.texi (-ffile-prefix-map): Document.
6982 (-fdebug-prefix-map): Update description.
6984 2018-01-18 Martin Liska <mliska@suse.cz>
6986 * config/i386/i386.c (indirect_thunk_name): Document that also
6988 (output_indirect_thunk): Document why both instructions
6989 (pause and lfence) are generated.
6991 2018-01-18 Richard Biener <rguenther@suse.de>
6993 PR tree-optimization/83887
6994 * graphite-scop-detection.c
6995 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6996 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6997 (scop_detection::merge_sese): Re-implement with a flood-fill
6998 algorithm that properly finds a SESE region if it exists.
7000 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7003 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7004 pointer_diff optimizations use view_convert instead of convert.
7006 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7008 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7009 Generate different code for -mno-speculate-indirect-jumps.
7010 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7011 (*call_indirect_aix<mode>): Disable for
7012 -mno-speculate-indirect-jumps.
7013 (*call_indirect_aix<mode>_nospec): New define_insn.
7014 (*call_value_indirect_aix<mode>): Disable for
7015 -mno-speculate-indirect-jumps.
7016 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7017 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7018 -mno-speculate-indirect-jumps.
7019 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7021 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7023 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7024 long double type, set the flags for noting the default long double
7025 type, even if we don't pass or return a long double type.
7027 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7030 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7033 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7035 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7036 support for merge[hl].
7037 (fold_mergehl_helper): New helper function.
7038 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7039 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7040 (altivec_vmrglw_direct): Add xxmrglw insn.
7042 2018-01-17 Andrew Waterman <andrew@sifive.com>
7044 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7045 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7047 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7050 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7051 call the lto_location_cache before reading the
7052 DECL_SOURCE_LOCATION of the types.
7054 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7055 Richard Sandiford <richard.sandiford@linaro.org>
7057 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7058 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7059 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7060 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7061 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7063 * config/aarch64/constraints.md (aarch64_movti_operand):
7065 * config/aarch64/predicates.md (Uti): Add new constraint.
7067 2018-01-17 Carl Love <cel@us.ibm.com>
7069 * config/rs6000/vsx.md (define_expand xl_len_r,
7070 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7071 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7073 (define_expand, define_insn): Move the shift left from the
7074 define_insn to the define_expand for lxvl and stxvl instructions.
7075 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7076 and XL_LEN_R definitions to PURE.
7078 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7080 * config/i386/i386.c (indirect_thunk_name): Declare regno
7081 as unsigned int. Compare regno with INVALID_REGNUM.
7082 (output_indirect_thunk): Ditto.
7083 (output_indirect_thunk_function): Ditto.
7084 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7085 in the call to output_indirect_thunk_function.
7087 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7090 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7091 rather than the size of inner_type to determine the stack slot size
7092 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7094 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7097 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7100 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7102 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7103 endian Linux systems to optionally enable multilibs for selecting
7104 the long double type if the user configured an explicit type.
7105 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7106 have no long double multilibs if not defined.
7107 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7108 warn if the user used -mabi={ieee,ibm}longdouble and we built
7109 multilibs for long double.
7110 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7111 appropriate multilib option.
7112 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7114 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7115 for building long double multilibs.
7116 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7118 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7120 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7123 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7125 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7128 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7131 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7134 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7136 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7137 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7139 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7141 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7142 different rtl trees depending on TARGET_64BIT.
7143 (rs6000_gen_lvx): Likewise.
7145 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7147 * config/visium/visium.md (nop): Tweak comment.
7148 (hazard_nop): Likewise.
7150 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7152 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7153 -mspeculate-indirect-jumps.
7154 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7155 for -mno-speculate-indirect-jumps.
7156 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7157 (*call_value_indirect_elfv2<mode>): Disable for
7158 -mno-speculate-indirect-jumps.
7159 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7160 (indirect_jump): Emit different RTL for
7161 -mno-speculate-indirect-jumps.
7162 (*indirect_jump<mode>): Disable for
7163 -mno-speculate-indirect-jumps.
7164 (*indirect_jump<mode>_nospec): New define_insn.
7165 (tablejump): Emit different RTL for
7166 -mno-speculate-indirect-jumps.
7167 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7168 (tablejumpsi_nospec): New define_expand.
7169 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7170 (tablejumpdi_nospec): New define_expand.
7171 (*tablejump<mode>_internal1): Disable for
7172 -mno-speculate-indirect-jumps.
7173 (*tablejump<mode>_internal1_nospec): New define_insn.
7174 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7177 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7179 * caller-save.c (insert_save): Drop unnecessary parameter. All
7182 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7183 Richard Biener <rguenth@suse.de>
7186 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7187 return early, inline manually is_gimple_sizepos. Make sure if we
7188 call gimplify_expr we don't end up with a gimple constant.
7189 * tree.c (variably_modified_type_p): Don't return true for
7190 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7191 * gimplify.h (is_gimple_sizepos): Remove.
7193 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7195 PR tree-optimization/83857
7196 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7197 vectorizable_live_operation for pure SLP statements.
7198 (vectorizable_live_operation): Handle PHIs.
7200 2018-01-16 Richard Biener <rguenther@suse.de>
7202 PR tree-optimization/83867
7203 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7204 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7206 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7209 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7210 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7211 If off is not INTEGER_CST, issue a may not be aligned warning
7212 rather than isn't aligned. Use isn%'t rather than isn't.
7213 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7215 <case MULT_EXPR>: Improve the case when bottom and one of the
7216 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7217 operand, in that case check if the other operand is multiple of
7218 bottom divided by the INTEGER_CST operand.
7220 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7223 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7224 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7225 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7226 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7227 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7228 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7229 (pa_function_arg_size): New function.
7231 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7233 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7234 in a separate statement.
7236 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7238 PR tree-optimization/83847
7239 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7240 group gathers and scatters.
7242 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7244 PR rtl-optimization/86620
7245 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7247 PR rtl-optimization/83213
7248 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7249 to last if both are JUMP_INSNs.
7251 PR tree-optimization/83843
7252 * gimple-ssa-store-merging.c
7253 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7254 store_immediate_info for bswap/nop orig_stores.
7256 2018-01-15 Andrew Waterman <andrew@sifive.com>
7258 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7260 <UDIV>: Increase cost if !TARGET_DIV.
7262 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7264 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7265 (define_attr "cr_logical_3op"): New.
7266 (cceq_ior_compare): Adjust.
7267 (cceq_ior_compare_complement): Adjust.
7268 (*cceq_rev_compare): Adjust.
7269 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7270 (is_cracked_insn): Adjust.
7271 (insn_must_be_first_in_group): Adjust.
7272 * config/rs6000/40x.md: Adjust.
7273 * config/rs6000/440.md: Adjust.
7274 * config/rs6000/476.md: Adjust.
7275 * config/rs6000/601.md: Adjust.
7276 * config/rs6000/603.md: Adjust.
7277 * config/rs6000/6xx.md: Adjust.
7278 * config/rs6000/7450.md: Adjust.
7279 * config/rs6000/7xx.md: Adjust.
7280 * config/rs6000/8540.md: Adjust.
7281 * config/rs6000/cell.md: Adjust.
7282 * config/rs6000/e300c2c3.md: Adjust.
7283 * config/rs6000/e500mc.md: Adjust.
7284 * config/rs6000/e500mc64.md: Adjust.
7285 * config/rs6000/e5500.md: Adjust.
7286 * config/rs6000/e6500.md: Adjust.
7287 * config/rs6000/mpc.md: Adjust.
7288 * config/rs6000/power4.md: Adjust.
7289 * config/rs6000/power5.md: Adjust.
7290 * config/rs6000/power6.md: Adjust.
7291 * config/rs6000/power7.md: Adjust.
7292 * config/rs6000/power8.md: Adjust.
7293 * config/rs6000/power9.md: Adjust.
7294 * config/rs6000/rs64.md: Adjust.
7295 * config/rs6000/titan.md: Adjust.
7297 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7299 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7300 ix86_indirect_branch_register logic.
7302 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7304 * config/i386/constraints.md (Bs): Update
7305 ix86_indirect_branch_register check. Don't check
7306 ix86_indirect_branch_register with GOT_memory_operand.
7308 * config/i386/predicates.md (GOT_memory_operand): Don't check
7309 ix86_indirect_branch_register here.
7310 (GOT32_symbol_operand): Likewise.
7312 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7314 * config/i386/predicates.md (constant_call_address_operand):
7315 Rewrite ix86_indirect_branch_register logic.
7316 (sibcall_insn_operand): Likewise.
7318 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7320 * config/i386/constraints.md (Bs): Replace
7321 ix86_indirect_branch_thunk_register with
7322 ix86_indirect_branch_register.
7324 * config/i386/i386.md (indirect_jump): Likewise.
7325 (tablejump): Likewise.
7326 (*sibcall_memory): Likewise.
7327 (*sibcall_value_memory): Likewise.
7328 Peepholes of indirect call and jump via memory: Likewise.
7329 * config/i386/i386.opt: Likewise.
7330 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7331 (GOT_memory_operand): Likewise.
7332 (call_insn_operand): Likewise.
7333 (sibcall_insn_operand): Likewise.
7334 (GOT32_symbol_operand): Likewise.
7336 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7339 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7340 type rather than type addr's type points to.
7341 (expand_omp_atomic_mutex): Likewise.
7342 (expand_omp_atomic): Likewise.
7344 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7347 * config/i386/i386.c (output_indirect_thunk_function): Use
7348 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7349 for __x86_return_thunk.
7351 2018-01-15 Richard Biener <rguenther@suse.de>
7354 * expmed.c (extract_bit_field_1): Fix typo.
7356 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7359 * config/arm/iterators.md (VF): New mode iterator.
7360 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7361 Remove integer-related logic from pattern.
7362 (neon_vabd<mode>_3): Likewise.
7364 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7367 * common.opt (fstrict-overflow): No longer an alias.
7368 (fwrapv-pointer): New option.
7369 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7370 also for pointer types based on flag_wrapv_pointer.
7371 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7372 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7373 opts->x_flag_wrapv got set.
7374 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7375 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7376 POINTER_TYPE_OVERFLOW_UNDEFINED.
7377 * match.pd: Likewise in address comparison pattern.
7378 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7380 2018-01-15 Richard Biener <rguenther@suse.de>
7383 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7384 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7385 Reset type names to their identifier if their TYPE_DECL doesn't
7386 have linkage (and thus is used for ODR and devirt).
7387 (save_debug_info_for_decl): Remove.
7388 (save_debug_info_for_type): Likewise.
7389 (add_tree_to_fld_list): Adjust.
7390 * tree-pretty-print.c (dump_generic_node): Make dumping of
7391 type names more robust.
7393 2018-01-15 Richard Biener <rguenther@suse.de>
7395 * BASE-VER: Bump to 8.0.1.
7397 2018-01-14 Martin Sebor <msebor@redhat.com>
7400 * builtins.c (check_access): Avoid warning when the no-warning bit
7403 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7405 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7406 * ira-color (allocno_hard_regs_compare): Likewise.
7408 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7411 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7412 Use .pushsection/.popsection.
7414 2018-01-14 Martin Sebor <msebor@redhat.com>
7417 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7419 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7421 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7422 entry from extra_headers.
7423 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7424 extra_headers, make the list bitwise identical to the i?86-*-* one.
7426 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7428 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7429 -mcmodel=large with -mindirect-branch=thunk,
7430 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7431 -mfunction-return=thunk-extern.
7432 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7433 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7434 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7436 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7438 * config/i386/i386.c (print_reg): Print the name of the full
7439 integer register without '%'.
7440 (ix86_print_operand): Handle 'V'.
7441 * doc/extend.texi: Document 'V' modifier.
7443 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7445 * config/i386/constraints.md (Bs): Disallow memory operand for
7446 -mindirect-branch-register.
7448 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7449 (GOT_memory_operand): Likewise.
7450 (call_insn_operand): Likewise.
7451 (sibcall_insn_operand): Likewise.
7452 (GOT32_symbol_operand): Likewise.
7453 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7454 for -mindirect-branch-register.
7455 (tablejump): Likewise.
7456 (*sibcall_memory): Likewise.
7457 (*sibcall_value_memory): Likewise.
7458 Disallow peepholes of indirect call and jump via memory for
7459 -mindirect-branch-register.
7460 (*call_pop): Replace m with Bw.
7461 (*call_value_pop): Likewise.
7462 (*sibcall_pop_memory): Replace m with Bs.
7463 * config/i386/i386.opt (mindirect-branch-register): New option.
7464 * doc/invoke.texi: Document -mindirect-branch-register option.
7466 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7468 * config/i386/i386-protos.h (ix86_output_function_return): New.
7469 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7470 set function_return_type.
7471 (indirect_thunk_name): Add ret_p to indicate thunk for function
7473 (output_indirect_thunk_function): Pass false to
7474 indirect_thunk_name.
7475 (ix86_output_indirect_branch_via_reg): Likewise.
7476 (ix86_output_indirect_branch_via_push): Likewise.
7477 (output_indirect_thunk_function): Create alias for function
7478 return thunk if regno < 0.
7479 (ix86_output_function_return): New function.
7480 (ix86_handle_fndecl_attribute): Handle function_return.
7481 (ix86_attribute_table): Add function_return.
7482 * config/i386/i386.h (machine_function): Add
7483 function_return_type.
7484 * config/i386/i386.md (simple_return_internal): Use
7485 ix86_output_function_return.
7486 (simple_return_internal_long): Likewise.
7487 * config/i386/i386.opt (mfunction-return=): New option.
7488 (indirect_branch): Mention -mfunction-return=.
7489 * doc/extend.texi: Document function_return function attribute.
7490 * doc/invoke.texi: Document -mfunction-return= option.
7492 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7494 * config/i386/i386-opts.h (indirect_branch): New.
7495 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7496 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7497 with local indirect jump when converting indirect call and jump.
7498 (ix86_set_indirect_branch_type): New.
7499 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7500 (indirectlabelno): New.
7501 (indirect_thunk_needed): Likewise.
7502 (indirect_thunk_bnd_needed): Likewise.
7503 (indirect_thunks_used): Likewise.
7504 (indirect_thunks_bnd_used): Likewise.
7505 (INDIRECT_LABEL): Likewise.
7506 (indirect_thunk_name): Likewise.
7507 (output_indirect_thunk): Likewise.
7508 (output_indirect_thunk_function): Likewise.
7509 (ix86_output_indirect_branch_via_reg): Likewise.
7510 (ix86_output_indirect_branch_via_push): Likewise.
7511 (ix86_output_indirect_branch): Likewise.
7512 (ix86_output_indirect_jmp): Likewise.
7513 (ix86_code_end): Call output_indirect_thunk_function if needed.
7514 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7516 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7517 (ix86_attribute_table): Add indirect_branch.
7518 * config/i386/i386.h (machine_function): Add indirect_branch_type
7519 and has_local_indirect_jump.
7520 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7522 (tablejump): Likewise.
7523 (*indirect_jump): Use ix86_output_indirect_jmp.
7524 (*tablejump_1): Likewise.
7525 (simple_return_indirect_internal): Likewise.
7526 * config/i386/i386.opt (mindirect-branch=): New option.
7527 (indirect_branch): New.
7530 (thunk-inline): Likewise.
7531 (thunk-extern): Likewise.
7532 * doc/extend.texi: Document indirect_branch function attribute.
7533 * doc/invoke.texi: Document -mindirect-branch= option.
7535 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7538 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7540 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7542 * ipa-inline.c (want_inline_small_function_p): Return false if
7543 inlining has already failed with CIF_FINAL_ERROR.
7544 (update_caller_keys): Call want_inline_small_function_p before
7546 (update_callee_keys): Likewise.
7548 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7550 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7552 (rs6000_quadword_masked_address_p): Likewise.
7553 (quad_aligned_load_p): Likewise.
7554 (quad_aligned_store_p): Likewise.
7555 (const_load_sequence_p): Add comment to describe the outer-most loop.
7556 (mimic_memory_attributes_and_flags): New function.
7557 (rs6000_gen_stvx): Likewise.
7558 (replace_swapped_aligned_store): Likewise.
7559 (rs6000_gen_lvx): Likewise.
7560 (replace_swapped_aligned_load): Likewise.
7561 (replace_swapped_load_constant): Capitalize argument name in
7562 comment describing this function.
7563 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7564 and stores that access quad-word aligned addresses and replace
7565 with stvx or lvx instructions when appropriate.
7566 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7567 New function prototype.
7568 (rs6000_quadword_masked_address_p): Likewise.
7569 (rs6000_gen_lvx): Likewise.
7570 (rs6000_gen_stvx): Likewise.
7571 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7572 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7573 when memory address is aligned.
7574 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7575 this split to select lvx instruction when memory address is aligned.
7576 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7577 instruction when memory address is aligned.
7578 (*vsx_le_perm_load_v16qi): Likewise.
7579 (four unnamed splitters): Modify to select the stvx instruction
7580 when memory is aligned.
7582 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7584 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7585 which appears in the queue multiple times.
7587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7588 Alan Hayward <alan.hayward@arm.com>
7589 David Sherwood <david.sherwood@arm.com>
7591 * tree-vectorizer.h (vec_lower_bound): New structure.
7592 (_loop_vec_info): Add check_nonzero and lower_bounds.
7593 (LOOP_VINFO_CHECK_NONZERO): New macro.
7594 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7595 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7596 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7597 fields. Make seg_len the distance travelled, not including the
7599 (dr_direction_indicator): Declare.
7600 (dr_zero_step_indicator): Likewise.
7601 (dr_known_forward_stride_p): Likewise.
7602 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7604 (runtime_alias_check_p): Allow runtime alias checks with
7606 (operator ==): Compare access_size and align.
7607 (prune_runtime_alias_test_list): Rework for new distinction between
7608 the access_size and seg_len.
7609 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7611 (get_segment_min_max): New function.
7612 (create_intersect_range_checks): Use it.
7613 (dr_step_indicator): New function.
7614 (dr_direction_indicator): Likewise.
7615 (dr_zero_step_indicator): Likewise.
7616 (dr_known_forward_stride_p): Likewise.
7617 * tree-loop-distribution.c (data_ref_segment_size): Return
7618 DR_STEP * (niters - 1).
7619 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7621 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7622 (vect_preserves_scalar_order_p): New function, split out from...
7623 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7624 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7625 (vect_vfa_access_size): New function.
7626 (vect_vfa_align): Likewise.
7627 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7628 (dump_lower_bound): New function.
7629 (vect_check_lower_bound): Likewise.
7630 (vect_small_gap_p): Likewise.
7631 (vectorizable_with_step_bound_p): Likewise.
7632 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7633 depencies if the vectorization factor is 1. Convert the checks
7634 for nonzero steps into checks on the bounds of DR_STEP. Try using
7635 a bunds check for variable steps if the minimum required step is
7636 relatively small. Update calls to the dr_with_seg_len
7637 constructor and to vect_compile_time_alias.
7638 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7640 (vect_loop_versioning): Call it.
7641 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7643 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7646 Alan Hayward <alan.hayward@arm.com>
7647 David Sherwood <david.sherwood@arm.com>
7649 * doc/sourcebuild.texi (vect_scatter_store): Document.
7650 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7652 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7654 * genopinit.c (main): Add supports_vec_scatter_store and
7655 supports_vec_scatter_store_cached to target_optabs.
7656 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7657 IFN_MASK_SCATTER_STORE.
7658 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7660 * internal-fn.h (internal_store_fn_p): Declare.
7661 (internal_fn_stored_value_index): Likewise.
7662 * internal-fn.c (scatter_store_direct): New macro.
7663 (expand_scatter_store_optab_fn): New function.
7664 (direct_scatter_store_optab_supported_p): New macro.
7665 (internal_store_fn_p): New function.
7666 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7667 IFN_MASK_SCATTER_STORE.
7668 (internal_fn_mask_index): Likewise.
7669 (internal_fn_stored_value_index): New function.
7670 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7672 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7673 * optabs-query.c (supports_vec_scatter_store_p): New function.
7674 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7675 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7676 true for scatter stores.
7677 (vect_gather_scatter_fn_p): Handle scatter stores too.
7678 (vect_check_gather_scatter): Consider using scatter stores if
7679 supports_vec_scatter_store_p.
7680 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7682 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7683 internal_fn_stored_value_index.
7684 (check_load_store_masking): Handle scatter stores too.
7685 (vect_get_store_rhs): Make public.
7686 (vectorizable_call): Use internal_store_fn_p.
7687 (vectorizable_store): Handle scatter store internal functions.
7688 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7689 when deciding whether the end of the group has been reached.
7690 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7691 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7692 (mask_scatter_store<mode>): New insns.
7694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7695 Alan Hayward <alan.hayward@arm.com>
7696 David Sherwood <david.sherwood@arm.com>
7698 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7699 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7700 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7702 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7703 Use vect_truncate_gather_scatter_offset if we can't treat the
7704 operation as a normal gather load or scatter store.
7705 (get_group_load_store_type): Take the gather_scatter_info
7706 as argument. Try using a gather load or scatter store for
7707 single-element groups.
7708 (get_load_store_type): Update calls to get_group_load_store_type
7709 and vect_use_strided_gather_scatters_p.
7711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7712 Alan Hayward <alan.hayward@arm.com>
7713 David Sherwood <david.sherwood@arm.com>
7715 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7716 optional tree argument.
7717 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7719 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7720 but continue to use the current value as a fallback.
7721 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7722 to compare the updates.
7723 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7724 (get_load_store_type): Use it when handling a strided access.
7725 (vect_get_strided_load_store_ops): New function.
7726 (vect_get_data_ptr_increment): Likewise.
7727 (vectorizable_load): Handle strided gather loads. Always pass
7728 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7731 Alan Hayward <alan.hayward@arm.com>
7732 David Sherwood <david.sherwood@arm.com>
7734 * doc/md.texi (gather_load@var{m}): Document.
7735 (mask_gather_load@var{m}): Likewise.
7736 * genopinit.c (main): Add supports_vec_gather_load and
7737 supports_vec_gather_load_cached to target_optabs.
7738 * optabs-tree.c (init_tree_optimization_optabs): Use
7739 ggc_cleared_alloc to allocate target_optabs.
7740 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7741 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7743 * internal-fn.h (internal_load_fn_p): Declare.
7744 (internal_gather_scatter_fn_p): Likewise.
7745 (internal_fn_mask_index): Likewise.
7746 (internal_gather_scatter_fn_supported_p): Likewise.
7747 * internal-fn.c (gather_load_direct): New macro.
7748 (expand_gather_load_optab_fn): New function.
7749 (direct_gather_load_optab_supported_p): New macro.
7750 (direct_internal_fn_optab): New function.
7751 (internal_load_fn_p): Likewise.
7752 (internal_gather_scatter_fn_p): Likewise.
7753 (internal_fn_mask_index): Likewise.
7754 (internal_gather_scatter_fn_supported_p): Likewise.
7755 * optabs-query.c (supports_at_least_one_mode_p): New function.
7756 (supports_vec_gather_load_p): Likewise.
7757 * optabs-query.h (supports_vec_gather_load_p): Declare.
7758 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7759 and memory_type field.
7760 (NUM_PATTERNS): Bump to 15.
7761 * tree-vect-data-refs.c: Include internal-fn.h.
7762 (vect_gather_scatter_fn_p): New function.
7763 (vect_describe_gather_scatter_call): Likewise.
7764 (vect_check_gather_scatter): Try using internal functions for
7765 gather loads. Recognize existing calls to a gather load function.
7766 (vect_analyze_data_refs): Consider using gather loads if
7767 supports_vec_gather_load_p.
7768 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7769 (vect_get_gather_scatter_offset_type): Likewise.
7770 (vect_convert_mask_for_vectype): Likewise.
7771 (vect_add_conversion_to_patterm): Likewise.
7772 (vect_try_gather_scatter_pattern): Likewise.
7773 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7774 (vect_vect_recog_func_ptrs): Add it.
7775 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7776 internal_fn_mask_index and internal_gather_scatter_fn_p.
7777 (check_load_store_masking): Take the gather_scatter_info as an
7778 argument and handle gather loads.
7779 (vect_get_gather_scatter_ops): New function.
7780 (vectorizable_call): Check internal_load_fn_p.
7781 (vectorizable_load): Likewise. Handle gather load internal
7783 (vectorizable_store): Update call to check_load_store_masking.
7784 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7785 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7786 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7787 (aarch64_gather_scale_operand_d): New predicates.
7788 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7789 (mask_gather_load<mode>): New insns.
7791 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7792 Alan Hayward <alan.hayward@arm.com>
7793 David Sherwood <david.sherwood@arm.com>
7795 * optabs.def (fold_left_plus_optab): New optab.
7796 * doc/md.texi (fold_left_plus_@var{m}): Document.
7797 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7798 * internal-fn.c (fold_left_direct): Define.
7799 (expand_fold_left_optab_fn): Likewise.
7800 (direct_fold_left_optab_supported_p): Likewise.
7801 * fold-const-call.c (fold_const_fold_left): New function.
7802 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7803 * tree-parloops.c (valid_reduction_p): New function.
7804 (gather_scalar_reductions): Use it.
7805 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7806 (vect_finish_replace_stmt): Declare.
7807 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7808 (needs_fold_left_reduction_p): New function, split out from...
7809 (vect_is_simple_reduction): ...here. Accept reductions that
7810 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7811 (vect_force_simple_reduction): Also store the reduction type in
7812 the assignment's STMT_VINFO_REDUC_TYPE.
7813 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7814 (merge_with_identity): New function.
7815 (vect_expand_fold_left): Likewise.
7816 (vectorize_fold_left_reduction): Likewise.
7817 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7818 scalar phi in place for it. Check for target support and reject
7819 cases that would reassociate the operation. Defer the transform
7820 phase to vectorize_fold_left_reduction.
7821 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7822 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7823 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7827 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7828 call to ifc_temp_var.
7830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7831 Alan Hayward <alan.hayward@arm.com>
7832 David Sherwood <david.sherwood@arm.com>
7834 * target.def (legitimize_address_displacement): Take the original
7835 offset as a poly_int.
7836 * targhooks.h (default_legitimize_address_displacement): Update
7838 * targhooks.c (default_legitimize_address_displacement): Likewise.
7839 * doc/tm.texi: Regenerate.
7840 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7841 as an argument, moving assert of ad->disp == ad->disp_term to...
7842 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7843 Try calling targetm.legitimize_address_displacement before expanding
7844 the address rather than afterwards, and adjust for the new interface.
7845 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7846 Match the new hook interface. Handle SVE addresses.
7847 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7852 * Makefile.in (OBJS): Add early-remat.o.
7853 * target.def (select_early_remat_modes): New hook.
7854 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7855 * doc/tm.texi: Regenerate.
7856 * targhooks.h (default_select_early_remat_modes): Declare.
7857 * targhooks.c (default_select_early_remat_modes): New function.
7858 * timevar.def (TV_EARLY_REMAT): New timevar.
7859 * passes.def (pass_early_remat): New pass.
7860 * tree-pass.h (make_pass_early_remat): Declare.
7861 * early-remat.c: New file.
7862 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7864 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7867 Alan Hayward <alan.hayward@arm.com>
7868 David Sherwood <david.sherwood@arm.com>
7870 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7871 vfm1 with a bound_epilog parameter.
7872 (vect_do_peeling): Update calls accordingly, and move the prologue
7873 call earlier in the function. Treat the base bound_epilog as 0 for
7874 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7875 this base when peeling for gaps.
7876 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7877 with fully-masked loops.
7878 (vect_estimate_min_profitable_iters): Handle the single peeled
7879 iteration in that case.
7881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7882 Alan Hayward <alan.hayward@arm.com>
7883 David Sherwood <david.sherwood@arm.com>
7885 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7886 single-element interleaving even if the size is not a power of 2.
7887 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7888 accesses for single-element interleaving if the group size is
7891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7892 Alan Hayward <alan.hayward@arm.com>
7893 David Sherwood <david.sherwood@arm.com>
7895 * doc/md.texi (fold_extract_last_@var{m}): Document.
7896 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7897 * optabs.def (fold_extract_last_optab): New optab.
7898 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7899 * internal-fn.c (fold_extract_direct): New macro.
7900 (expand_fold_extract_optab_fn): Likewise.
7901 (direct_fold_extract_optab_supported_p): Likewise.
7902 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7903 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7904 EXTRACT_LAST_REDUCTION.
7905 (get_initial_def_for_reduction): Do not create an initial vector
7906 for EXTRACT_LAST_REDUCTION reductions.
7907 (vectorizable_reduction): Leave the scalar phi in place for
7908 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7909 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7910 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7911 transform phase to vectorizable_condition.
7912 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7914 (vect_finish_stmt_generation): ...here.
7915 (vect_finish_replace_stmt): New function.
7916 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7917 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7919 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7922 Alan Hayward <alan.hayward@arm.com>
7923 David Sherwood <david.sherwood@arm.com>
7925 * doc/md.texi (extract_last_@var{m}): Document.
7926 * optabs.def (extract_last_optab): New optab.
7927 * internal-fn.def (EXTRACT_LAST): New internal function.
7928 * internal-fn.c (cond_unary_direct): New macro.
7929 (expand_cond_unary_optab_fn): Likewise.
7930 (direct_cond_unary_optab_supported_p): Likewise.
7931 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7932 loops using EXTRACT_LAST.
7933 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7934 (extract_last_<mode>): ...this optab.
7935 (vec_extract<mode><Vel>): Update accordingly.
7937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7938 Alan Hayward <alan.hayward@arm.com>
7939 David Sherwood <david.sherwood@arm.com>
7941 * target.def (empty_mask_is_expensive): New hook.
7942 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7943 * doc/tm.texi: Regenerate.
7944 * targhooks.h (default_empty_mask_is_expensive): Declare.
7945 * targhooks.c (default_empty_mask_is_expensive): New function.
7946 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7947 if the target says that empty masks are expensive.
7948 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7950 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7953 Alan Hayward <alan.hayward@arm.com>
7954 David Sherwood <david.sherwood@arm.com>
7956 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7957 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7958 (vect_use_loop_mask_for_alignment_p): New function.
7959 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7960 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7961 niters_skip argument. Make sure that the first niters_skip elements
7962 of the first iteration are inactive.
7963 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7964 Update call to vect_set_loop_masks_directly.
7965 (get_misalign_in_elems): New function, split out from...
7966 (vect_gen_prolog_loop_niters): ...here.
7967 (vect_update_init_of_dr): Take a code argument that specifies whether
7968 the adjustment should be added or subtracted.
7969 (vect_update_init_of_drs): Likewise.
7970 (vect_prepare_for_masked_peels): New function.
7971 (vect_do_peeling): Skip prologue peeling if we're using a mask
7972 instead. Update call to vect_update_inits_of_drs.
7973 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7975 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7976 alignment. Do not include the number of peeled iterations in
7977 the minimum threshold in that case.
7978 (vectorizable_induction): Adjust the start value down by
7979 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7980 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7981 Take the number of skipped iterations into account when calculating
7983 * tree-vect-stmts.c (vect_gen_while_not): New function.
7985 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7986 Alan Hayward <alan.hayward@arm.com>
7987 David Sherwood <david.sherwood@arm.com>
7989 * doc/sourcebuild.texi (vect_fully_masked): Document.
7990 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7992 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7994 (vect_analyze_loop_2): ...here. Don't check the vectorization
7995 factor against the number of loop iterations if the loop is
7998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7999 Alan Hayward <alan.hayward@arm.com>
8000 David Sherwood <david.sherwood@arm.com>
8002 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8003 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8004 (dump_groups): Update accordingly.
8005 (iv_use::mem_type): New member variable.
8006 (address_p): New function.
8007 (record_use): Add a mem_type argument and initialize the new
8009 (record_group_use): Add a mem_type argument. Use address_p.
8010 Remove obsolete null checks of base_object. Update call to record_use.
8011 (find_interesting_uses_op): Update call to record_group_use.
8012 (find_interesting_uses_cond): Likewise.
8013 (find_interesting_uses_address): Likewise.
8014 (get_mem_type_for_internal_fn): New function.
8015 (find_address_like_use): Likewise.
8016 (find_interesting_uses_stmt): Try find_address_like_use before
8017 calling find_interesting_uses_op.
8018 (addr_offset_valid_p): Use the iv mem_type field as the type
8019 of the addressed memory.
8020 (add_autoinc_candidates): Likewise.
8021 (get_address_cost): Likewise.
8022 (split_small_address_groups_p): Use address_p.
8023 (split_address_groups): Likewise.
8024 (add_iv_candidate_for_use): Likewise.
8025 (autoinc_possible_for_pair): Likewise.
8026 (rewrite_groups): Likewise.
8027 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8028 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8029 (get_alias_ptr_type_for_ptr_address): New function.
8030 (rewrite_use_address): Rewrite address uses in calls that were
8031 identified by find_address_like_use.
8033 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8034 Alan Hayward <alan.hayward@arm.com>
8035 David Sherwood <david.sherwood@arm.com>
8037 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8039 * gimple-expr.h (is_gimple_addressable: Likewise.
8040 * gimple-expr.c (is_gimple_address): Likewise.
8041 * internal-fn.c (expand_call_mem_ref): New function.
8042 (expand_mask_load_optab_fn): Use it.
8043 (expand_mask_store_optab_fn): Likewise.
8045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8046 Alan Hayward <alan.hayward@arm.com>
8047 David Sherwood <david.sherwood@arm.com>
8049 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8050 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8051 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8052 (cond_umax@var{mode}): Document.
8053 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8054 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8055 (cond_umin_optab, cond_umax_optab): New optabs.
8056 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8057 (COND_IOR, COND_XOR): New internal functions.
8058 * internal-fn.h (get_conditional_internal_fn): Declare.
8059 * internal-fn.c (cond_binary_direct): New macro.
8060 (expand_cond_binary_optab_fn): Likewise.
8061 (direct_cond_binary_optab_supported_p): Likewise.
8062 (get_conditional_internal_fn): New function.
8063 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8064 Cope with reduction statements that are vectorized as calls rather
8066 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8067 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8068 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8069 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8070 (UNSPEC_COND_EOR): New unspecs.
8071 (optab): Add mappings for them.
8072 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8073 (sve_int_op, sve_fp_op): New int attributes.
8075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8076 Alan Hayward <alan.hayward@arm.com>
8077 David Sherwood <david.sherwood@arm.com>
8079 * optabs.def (while_ult_optab): New optab.
8080 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8081 * internal-fn.def (WHILE_ULT): New internal function.
8082 * internal-fn.h (direct_internal_fn_supported_p): New override
8083 that takes two types as argument.
8084 * internal-fn.c (while_direct): New macro.
8085 (expand_while_optab_fn): New function.
8086 (convert_optab_supported_p): Likewise.
8087 (direct_while_optab_supported_p): New macro.
8088 * wide-int.h (wi::udiv_ceil): New function.
8089 * tree-vectorizer.h (rgroup_masks): New structure.
8090 (vec_loop_masks): New typedef.
8091 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8093 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8094 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8095 (vect_max_vf): New function.
8096 (slpeel_make_loop_iterate_ntimes): Delete.
8097 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8098 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8099 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8100 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8101 internal-fn.h, stor-layout.h and optabs-query.h.
8102 (vect_set_loop_mask): New function.
8103 (add_preheader_seq): Likewise.
8104 (add_header_seq): Likewise.
8105 (interleave_supported_p): Likewise.
8106 (vect_maybe_permute_loop_masks): Likewise.
8107 (vect_set_loop_masks_directly): Likewise.
8108 (vect_set_loop_condition_masked): Likewise.
8109 (vect_set_loop_condition_unmasked): New function, split out from
8110 slpeel_make_loop_iterate_ntimes.
8111 (slpeel_make_loop_iterate_ntimes): Rename to..
8112 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8113 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8114 (vect_do_peeling): Update call accordingly.
8115 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8117 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8118 mask_compare_type, can_fully_mask_p and fully_masked_p.
8119 (release_vec_loop_masks): New function.
8120 (_loop_vec_info): Use it to free the loop masks.
8121 (can_produce_all_loop_masks_p): New function.
8122 (vect_get_max_nscalars_per_iter): Likewise.
8123 (vect_verify_full_masking): Likewise.
8124 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8125 retries, and free the mask rgroups before retrying. Check loop-wide
8126 reasons for disallowing fully-masked loops. Make the final decision
8127 about whether use a fully-masked loop or not.
8128 (vect_estimate_min_profitable_iters): Do not assume that peeling
8129 for the number of iterations will be needed for fully-masked loops.
8130 (vectorizable_reduction): Disable fully-masked loops.
8131 (vectorizable_live_operation): Likewise.
8132 (vect_halve_mask_nunits): New function.
8133 (vect_double_mask_nunits): Likewise.
8134 (vect_record_loop_mask): Likewise.
8135 (vect_get_loop_mask): Likewise.
8136 (vect_transform_loop): Handle the case in which the final loop
8137 iteration might handle a partial vector. Call vect_set_loop_condition
8138 instead of slpeel_make_loop_iterate_ntimes.
8139 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8140 (check_load_store_masking): New function.
8141 (prepare_load_store_mask): Likewise.
8142 (vectorizable_store): Handle fully-masked loops.
8143 (vectorizable_load): Likewise.
8144 (supportable_widening_operation): Use vect_halve_mask_nunits for
8146 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8147 (vect_gen_while): New function.
8148 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8149 (aarch64_uqdec<mode>): New insn.
8151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8152 Alan Hayward <alan.hayward@arm.com>
8153 David Sherwood <david.sherwood@arm.com>
8155 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8156 (reduc_xor_scal_optab): New optabs.
8157 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8158 (reduc_xor_scal_@var{m}): Document.
8159 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8160 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8162 * fold-const-call.c (fold_const_call): Handle them.
8163 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8164 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8165 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8166 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8167 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8168 (UNSPEC_XORV): New unspecs.
8169 (optab): Add entries for them.
8170 (BITWISEV): New int iterator.
8171 (bit_reduc_op): New int attributes.
8173 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8174 Alan Hayward <alan.hayward@arm.com>
8175 David Sherwood <david.sherwood@arm.com>
8177 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8178 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8179 * optabs.def (vec_shl_insert_optab): New optab.
8180 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8181 (duplicate_and_interleave): Likewise.
8182 * tree-vect-loop.c: Include internal-fn.h.
8183 (neutral_op_for_slp_reduction): New function, split out from
8184 get_initial_defs_for_reduction.
8185 (get_initial_def_for_reduction): Handle option 2 for variable-length
8186 vectors by loading the neutral value into a vector and then shifting
8187 the initial value into element 0.
8188 (get_initial_defs_for_reduction): Replace the code argument with
8189 the neutral value calculated by neutral_op_for_slp_reduction.
8190 Use gimple_build_vector for constant-length vectors.
8191 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8192 but the first group_size elements have a neutral value.
8193 Use duplicate_and_interleave otherwise.
8194 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8195 Update call to get_initial_defs_for_reduction. Handle SLP
8196 reductions for variable-length vectors by creating one vector
8197 result for each scalar result, with the elements associated
8198 with other scalar results stubbed out with the neutral value.
8199 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8200 Require IFN_VEC_SHL_INSERT for double reductions on
8201 variable-length vectors, or SLP reductions that have
8202 a neutral value. Require can_duplicate_and_interleave_p
8203 support for variable-length unchained SLP reductions if there
8204 is no neutral value, such as for MIN/MAX reductions. Also require
8205 the number of vector elements to be a multiple of the number of
8206 SLP statements when doing variable-length unchained SLP reductions.
8207 Update call to vect_create_epilog_for_reduction.
8208 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8209 and remove initial values.
8210 (duplicate_and_interleave): Make public.
8211 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8212 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8215 Alan Hayward <alan.hayward@arm.com>
8216 David Sherwood <david.sherwood@arm.com>
8218 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8219 (can_duplicate_and_interleave_p): New function.
8220 (vect_get_and_check_slp_defs): Take the vector of statements
8221 rather than just the current one. Remove excess parentheses.
8222 Restriction rejectinon of vect_constant_def and vect_external_def
8223 for variable-length vectors to boolean types, or types for which
8224 can_duplicate_and_interleave_p is false.
8225 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8226 (duplicate_and_interleave): New function.
8227 (vect_get_constant_vectors): Use gimple_build_vector for
8228 constant-length vectors and suitable variable-length constant
8229 vectors. Use duplicate_and_interleave for other variable-length
8230 vectors. Don't defer the update when inserting new statements.
8232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8233 Alan Hayward <alan.hayward@arm.com>
8234 David Sherwood <david.sherwood@arm.com>
8236 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8237 min_profitable_iters doesn't go negative.
8239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8240 Alan Hayward <alan.hayward@arm.com>
8241 David Sherwood <david.sherwood@arm.com>
8243 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8244 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8245 * optabs.def (vec_mask_load_lanes_optab): New optab.
8246 (vec_mask_store_lanes_optab): Likewise.
8247 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8248 (MASK_STORE_LANES): Likewise.
8249 * internal-fn.c (mask_load_lanes_direct): New macro.
8250 (mask_store_lanes_direct): Likewise.
8251 (expand_mask_load_optab_fn): Handle masked operations.
8252 (expand_mask_load_lanes_optab_fn): New macro.
8253 (expand_mask_store_optab_fn): Handle masked operations.
8254 (expand_mask_store_lanes_optab_fn): New macro.
8255 (direct_mask_load_lanes_optab_supported_p): Likewise.
8256 (direct_mask_store_lanes_optab_supported_p): Likewise.
8257 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8259 (vect_load_lanes_supported): Likewise.
8260 * tree-vect-data-refs.c (strip_conversion): New function.
8261 (can_group_stmts_p): Likewise.
8262 (vect_analyze_data_ref_accesses): Use it instead of checking
8263 for a pair of assignments.
8264 (vect_store_lanes_supported): Take a masked_p parameter.
8265 (vect_load_lanes_supported): Likewise.
8266 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8267 vect_store_lanes_supported and vect_load_lanes_supported.
8268 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8269 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8270 parameter. Don't allow gaps for masked accesses.
8271 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8272 and vect_load_lanes_supported.
8273 (get_load_store_type): Take a masked_p parameter and update
8274 call to get_group_load_store_type.
8275 (vectorizable_store): Update call to get_load_store_type.
8276 Handle IFN_MASK_STORE_LANES.
8277 (vectorizable_load): Update call to get_load_store_type.
8278 Handle IFN_MASK_LOAD_LANES.
8280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8281 Alan Hayward <alan.hayward@arm.com>
8282 David Sherwood <david.sherwood@arm.com>
8284 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8286 * config/aarch64/aarch64-protos.h
8287 (aarch64_sve_struct_memory_operand_p): Declare.
8288 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8289 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8290 (VPRED, vpred): Handle SVE structure modes.
8291 * config/aarch64/constraints.md (Utx): New constraint.
8292 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8293 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8294 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8295 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8296 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8297 structure modes. Split into pieces after RA.
8298 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8299 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8301 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8302 SVE structure modes.
8303 (aarch64_classify_address): Likewise.
8304 (sizetochar): Move earlier in file.
8305 (aarch64_print_operand): Handle SVE register lists.
8306 (aarch64_array_mode): New function.
8307 (aarch64_sve_struct_memory_operand_p): Likewise.
8308 (TARGET_ARRAY_MODE): Redefine.
8310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8311 Alan Hayward <alan.hayward@arm.com>
8312 David Sherwood <david.sherwood@arm.com>
8314 * target.def (array_mode): New target hook.
8315 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8316 * doc/tm.texi: Regenerate.
8317 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8318 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8319 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8321 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8325 Alan Hayward <alan.hayward@arm.com>
8326 David Sherwood <david.sherwood@arm.com>
8328 * fold-const.c (fold_binary_loc): Check the argument types
8329 rather than the result type when testing for a vector operation.
8331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8333 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8334 * doc/tm.texi: Regenerate.
8336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8337 Alan Hayward <alan.hayward@arm.com>
8338 David Sherwood <david.sherwood@arm.com>
8340 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8341 (sve): Document new AArch64 extension.
8342 * doc/md.texi (w): Extend the description of the AArch64
8343 constraint to include SVE vectors.
8344 (Upl, Upa): Document new AArch64 predicate constraints.
8345 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8347 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8348 (msve-vector-bits=): New option.
8349 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8350 SVE when these are disabled.
8351 (sve): New extension.
8352 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8353 modes. Adjust their number of units based on aarch64_sve_vg.
8354 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8355 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8356 aarch64_addr_query_type.
8357 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8358 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8359 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8360 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8361 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8362 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8363 (aarch64_simd_imm_zero_p): Delete.
8364 (aarch64_check_zero_based_sve_index_immediate): Declare.
8365 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8366 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8367 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8368 (aarch64_sve_float_mul_immediate_p): Likewise.
8369 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8371 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8372 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8373 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8374 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8375 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8376 (aarch64_regmode_natural_size): Likewise.
8377 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8378 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8380 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8381 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8382 for VG and the SVE predicate registers.
8383 (V_ALIASES): Add a "z"-prefixed alias.
8384 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8385 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8386 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8387 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8388 (REG_CLASS_NAMES): Add entries for them.
8389 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8390 and the predicate registers.
8391 (aarch64_sve_vg): Declare.
8392 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8393 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8394 (REGMODE_NATURAL_SIZE): Define.
8395 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8397 * config/aarch64/aarch64.c: Include cfgrtl.h.
8398 (simd_immediate_info): Add a constructor for series vectors,
8399 and an associated step field.
8400 (aarch64_sve_vg): New variable.
8401 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8402 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8403 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8404 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8405 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8406 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8407 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8408 (aarch64_get_mask_mode): New functions.
8409 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8410 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8411 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8412 predicate modes and predicate registers. Explicitly restrict
8413 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8414 to store a vector mode if it is recognized by
8415 aarch64_classify_vector_mode.
8416 (aarch64_regmode_natural_size): New function.
8417 (aarch64_hard_regno_caller_save_mode): Return the original mode
8419 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8420 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8421 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8422 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8424 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8425 does not overlap dest if the function is frame-related. Handle
8427 (aarch64_split_add_offset): New function.
8428 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8429 them aarch64_add_offset.
8430 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8431 and update call to aarch64_sub_sp.
8432 (aarch64_add_cfa_expression): New function.
8433 (aarch64_expand_prologue): Pass extra temporary registers to the
8434 functions above. Handle the case in which we need to emit new
8435 DW_CFA_expressions for registers that were originally saved
8436 relative to the stack pointer, but now have to be expressed
8437 relative to the frame pointer.
8438 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8440 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8441 IP0 and IP1 values for SVE frames.
8442 (aarch64_expand_vec_series): New function.
8443 (aarch64_expand_sve_widened_duplicate): Likewise.
8444 (aarch64_expand_sve_const_vector): Likewise.
8445 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8446 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8447 into the register, rather than emitting a SET directly.
8448 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8449 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8450 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8451 (offset_9bit_signed_scaled_p): New functions.
8452 (aarch64_replicate_bitmask_imm): New function.
8453 (aarch64_bitmask_imm): Use it.
8454 (aarch64_cannot_force_const_mem): Reject expressions involving
8455 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8456 (aarch64_classify_index): Handle SVE indices, by requiring
8457 a plain register index with a scale that matches the element size.
8458 (aarch64_classify_address): Handle SVE addresses. Assert that
8459 the mode of the address is VOIDmode or an integer mode.
8460 Update call to aarch64_classify_symbol.
8461 (aarch64_classify_symbolic_expression): Update call to
8462 aarch64_classify_symbol.
8463 (aarch64_const_vec_all_in_range_p): New function.
8464 (aarch64_print_vector_float_operand): Likewise.
8465 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8466 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8467 and the FP immediates 1.0 and 0.5.
8468 (aarch64_print_address_internal): Handle SVE addresses.
8469 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8470 (aarch64_regno_regclass): Handle predicate registers.
8471 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8473 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8474 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8475 (aarch64_convert_sve_vector_bits): New function.
8476 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8477 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8479 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8480 Handle SVE vector and predicate modes. Accept VL-based constants
8481 that need only one temporary register, and VL offsets that require
8482 no temporary registers.
8483 (aarch64_conditional_register_usage): Mark the predicate registers
8484 as fixed if SVE isn't available.
8485 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8486 Return true for SVE vector and predicate modes.
8487 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8488 rather than an unsigned int. Handle SVE modes.
8489 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8491 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8493 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8494 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8495 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8496 (aarch64_sve_float_mul_immediate_p): New functions.
8497 (aarch64_sve_valid_immediate): New function.
8498 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8499 Explicitly reject structure modes. Check for INDEX constants.
8500 Handle PTRUE and PFALSE constants.
8501 (aarch64_check_zero_based_sve_index_immediate): New function.
8502 (aarch64_simd_imm_zero_p): Delete.
8503 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8504 vector modes. Accept constants in the range of CNT[BHWD].
8505 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8506 ask for an Advanced SIMD mode.
8507 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8508 (aarch64_simd_vector_alignment): Handle SVE predicates.
8509 (aarch64_vectorize_preferred_vector_alignment): New function.
8510 (aarch64_simd_vector_alignment_reachable): Use it instead of
8512 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8513 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8515 (MAX_VECT_LEN): Delete.
8516 (expand_vec_perm_d): Add a vec_flags field.
8517 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8518 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8519 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8521 (aarch64_evpc_rev): Rename to...
8522 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8523 (aarch64_evpc_rev_global): New function.
8524 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8525 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8527 (aarch64_evpc_sve_tbl): New function.
8528 (aarch64_expand_vec_perm_const_1): Update after rename of
8529 aarch64_evpc_rev. Handle SVE permutes too, trying
8530 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8531 than aarch64_evpc_tbl.
8532 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8533 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8534 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8535 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8536 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8537 (aarch64_expand_sve_vcond): New functions.
8538 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8539 of aarch64_vector_mode_p.
8540 (aarch64_dwarf_poly_indeterminate_value): New function.
8541 (aarch64_compute_pressure_classes): Likewise.
8542 (aarch64_can_change_mode_class): Likewise.
8543 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8544 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8545 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8546 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8547 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8548 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8549 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8550 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8552 (Dn, Dl, Dr): Accept const as well as const_vector.
8553 (Dz): Likewise. Compare against CONST0_RTX.
8554 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8555 of "vector" where appropriate.
8556 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8557 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8558 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8559 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8560 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8561 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8562 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8563 (v_int_equiv): Extend to SVE modes.
8564 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8566 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8567 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8568 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8569 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8570 (SVE_COND_FP_CMP): New int iterators.
8571 (perm_hilo): Handle the new unpack unspecs.
8572 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8574 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8575 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8576 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8577 (aarch64_equality_operator, aarch64_constant_vector_operand)
8578 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8579 (aarch64_sve_nonimmediate_operand): Likewise.
8580 (aarch64_sve_general_operand): Likewise.
8581 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8582 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8583 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8584 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8585 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8586 (aarch64_sve_float_arith_immediate): Likewise.
8587 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8588 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8589 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8590 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8591 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8592 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8593 (aarch64_sve_float_arith_operand): Likewise.
8594 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8595 (aarch64_sve_float_mul_operand): Likewise.
8596 (aarch64_sve_vec_perm_operand): Likewise.
8597 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8598 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8599 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8600 as well as const_vector.
8601 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8602 in file. Use CONST0_RTX and CONSTM1_RTX.
8603 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8604 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8605 Use aarch64_simd_imm_zero.
8606 * config/aarch64/aarch64-sve.md: New file.
8607 * config/aarch64/aarch64.md: Include it.
8608 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8609 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8610 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8611 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8612 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8613 (sve): New attribute.
8614 (enabled): Disable instructions with the sve attribute unless
8616 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8617 aarch64_expand_mov_immediate.
8618 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8619 CNT[BHSD] immediates.
8620 (movti): Split CONST_POLY_INT moves into two halves.
8621 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8622 Split additions that need a temporary here if the destination
8623 is the stack pointer.
8624 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8625 (*add<mode>3_poly_1): New instruction.
8626 (set_clobber_cc): New expander.
8628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8630 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8631 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8632 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8633 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8634 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8635 Change innermode from fixed_mode_size to machine_mode.
8636 (simplify_subreg): Update call accordingly. Handle a constant-sized
8637 subreg of a variable-length CONST_VECTOR.
8639 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8640 Alan Hayward <alan.hayward@arm.com>
8641 David Sherwood <david.sherwood@arm.com>
8643 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8644 (add_offset_to_base): New function, split out from...
8645 (create_mem_ref): ...here. When handling a scale other than 1,
8646 check first whether the address is valid without the offset.
8647 Add it into the base if so, leaving the index and scale as-is.
8649 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8652 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8653 fold_for_warn before checking if arg2 is INTEGER_CST.
8655 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8657 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8658 (store_multiple_operation): Delete.
8659 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8660 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8661 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8662 guarded by TARGET_STRING.
8663 (rs6000_output_load_multiple): Delete.
8664 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8665 OPTION_MASK_STRING / TARGET_STRING handling.
8666 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8667 (const rs6000_opt_masks) <"string">: Change mask to 0.
8668 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8669 (MASK_STRING): Delete.
8670 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8672 (load_multiple): Delete.
8679 (store_multiple): Delete.
8686 (movmemsi_8reg): Delete.
8687 (corresponding unnamed define_insn): Delete.
8688 (movmemsi_6reg): Delete.
8689 (corresponding unnamed define_insn): Delete.
8690 (movmemsi_4reg): Delete.
8691 (corresponding unnamed define_insn): Delete.
8692 (movmemsi_2reg): Delete.
8693 (corresponding unnamed define_insn): Delete.
8694 (movmemsi_1reg): Delete.
8695 (corresponding unnamed define_insn): Delete.
8696 * config/rs6000/rs6000.opt (mno-string): New.
8697 (mstring): Replace by deprecation warning stub.
8698 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8700 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8702 * regrename.c (regrename_do_replace): If replacing the same
8703 reg multiple times, try to reuse last created gen_raw_REG.
8706 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8707 main to workaround a bug in GDB.
8709 2018-01-12 Tom de Vries <tom@codesourcery.com>
8712 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8714 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8716 PR rtl-optimization/80481
8717 * ira-color.c (get_cap_member): New function.
8718 (allocnos_conflict_by_live_ranges_p): Use it.
8719 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8720 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8722 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8725 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8726 (*saddl_se_1): Ditto.
8728 (*ssubl_se_1): Ditto.
8730 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8732 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8733 rather than wi::to_widest for DR_INITs.
8734 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8735 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8736 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8738 (vect_analyze_group_access_1): Note that here.
8740 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8742 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8743 polynomial type sizes.
8745 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8747 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8748 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8749 (gimple_add_tmp_var): Likewise.
8751 2018-01-12 Martin Liska <mliska@suse.cz>
8753 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8754 (gimple_alloc_sizes): Likewise.
8755 (dump_gimple_statistics): Use PRIu64 in printf format.
8756 * gimple.h: Change uint64_t to int.
8758 2018-01-12 Martin Liska <mliska@suse.cz>
8760 * tree-core.h: Use uint64_t instead of int.
8761 * tree.c (tree_node_counts): Likewise.
8762 (tree_node_sizes): Likewise.
8763 (dump_tree_statistics): Use PRIu64 in printf format.
8765 2018-01-12 Martin Liska <mliska@suse.cz>
8767 * Makefile.in: As qsort_chk is implemented in vec.c, add
8768 vec.o to linkage of gencfn-macros.
8769 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8770 passing the info to record_node_allocation_statistics.
8771 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8773 * ggc-common.c (struct ggc_usage): Add operator== and use
8774 it in operator< and compare function.
8775 * mem-stats.h (struct mem_usage): Likewise.
8776 * vec.c (struct vec_usage): Remove operator< and compare
8777 function. Can be simply inherited.
8779 2018-01-12 Martin Jambor <mjambor@suse.cz>
8782 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8783 * tree-ssa-math-opts.c: Include domwalk.h.
8784 (convert_mult_to_fma_1): New function.
8785 (fma_transformation_info): New type.
8786 (fma_deferring_state): Likewise.
8787 (cancel_fma_deferring): New function.
8788 (result_of_phi): Likewise.
8789 (last_fma_candidate_feeds_initial_phi): Likewise.
8790 (convert_mult_to_fma): Added deferring logic, split actual
8791 transformation to convert_mult_to_fma_1.
8792 (math_opts_dom_walker): New type.
8793 (math_opts_dom_walker::after_dom_children): New method, body moved
8794 here from pass_optimize_widening_mul::execute, added deferring logic
8796 (pass_optimize_widening_mul::execute): Moved most of code to
8797 math_opts_dom_walker::after_dom_children.
8798 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8799 * config/i386/i386.c (ix86_option_override_internal): Added
8800 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8802 2018-01-12 Richard Biener <rguenther@suse.de>
8805 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8806 inline instance vars.
8808 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8811 * config/rx/rx.c (rx_is_restricted_memory_address):
8814 2018-01-12 Richard Biener <rguenther@suse.de>
8816 PR tree-optimization/80846
8817 * target.def (split_reduction): New target hook.
8818 * targhooks.c (default_split_reduction): New function.
8819 * targhooks.h (default_split_reduction): Declare.
8820 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8821 target requests first reduce vectors by combining low and high
8823 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8824 (get_vectype_for_scalar_type_and_size): Export.
8825 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8826 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8827 * doc/tm.texi: Regenerate.
8828 * config/i386/i386.c (ix86_split_reduction): Implement
8829 TARGET_VECTORIZE_SPLIT_REDUCTION.
8831 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8834 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8835 in PIC mode except for TARGET_VXWORKS_RTP.
8836 * config/sparc/sparc.c: Include cfgrtl.h.
8837 (TARGET_INIT_PIC_REG): Define.
8838 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8839 (sparc_pic_register_p): New predicate.
8840 (sparc_legitimate_address_p): Use it.
8841 (sparc_legitimize_pic_address): Likewise.
8842 (sparc_delegitimize_address): Likewise.
8843 (sparc_mode_dependent_address_p): Likewise.
8844 (gen_load_pcrel_sym): Remove 4th parameter.
8845 (load_got_register): Adjust call to above. Remove obsolete stuff.
8846 (sparc_expand_prologue): Do not call load_got_register here.
8847 (sparc_flat_expand_prologue): Likewise.
8848 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8849 (sparc_use_pseudo_pic_reg): New function.
8850 (sparc_init_pic_reg): Likewise.
8851 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8852 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8854 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8856 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8857 Add item for branch_cost.
8859 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8861 PR rtl-optimization/83565
8862 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8863 not extend the result to a larger mode for rotate operations.
8864 (num_sign_bit_copies1): Likewise.
8866 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8869 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8871 Use values-Xc.o for -pedantic.
8872 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8874 2018-01-12 Martin Liska <mliska@suse.cz>
8877 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8879 (possible_polymorphic_call_targets): Use it.
8880 (ipa_devirt): Likewise.
8882 2018-01-12 Martin Liska <mliska@suse.cz>
8884 * profile-count.h (enum profile_quality): Use 0 as invalid
8885 enum value of profile_quality.
8887 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8889 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8890 -mext-string options.
8892 2018-01-12 Richard Biener <rguenther@suse.de>
8894 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8895 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8896 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8898 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8900 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8902 * configure.ac (--with-long-double-format): Add support for the
8903 configuration option to change the default long double format on
8905 * config.gcc (powerpc*-linux*-*): Likewise.
8906 * configure: Regenerate.
8907 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8908 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8909 used without modification.
8911 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8913 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8914 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8915 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8916 MISC_BUILTIN_SPEC_BARRIER.
8917 (rs6000_init_builtins): Likewise.
8918 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8920 (speculation_barrier): New define_insn.
8921 * doc/extend.texi: Document __builtin_speculation_barrier.
8923 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8926 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8927 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8928 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8930 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8931 integral modes instead of "ss" and "sd".
8932 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8933 vectors with 32-bit and 64-bit elements.
8934 (vecdupssescalarmodesuffix): New mode attribute.
8935 (vec_dup<mode>): Use it.
8937 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8940 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8941 frame if argument is passed on stack.
8943 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8946 * ree.c (combine_reaching_defs): Optimize also
8947 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8948 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8950 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8953 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8955 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8958 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8959 after they are computed.
8961 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8963 PR tree-optimization/83695
8964 * gimple-loop-linterchange.cc
8965 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8966 reset cached scev information after interchange.
8967 (pass_linterchange::execute): Remove call to scev_reset_htab.
8969 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8971 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8972 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8973 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8974 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8975 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8976 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8977 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8978 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8979 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8980 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8981 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8982 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8983 (V_lane_reg): Likewise.
8984 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8986 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8987 (vfmal_lane_low<mode>_intrinsic,
8988 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8989 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8990 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8991 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8992 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8993 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8995 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8997 * config/arm/arm-cpus.in (fp16fml): New feature.
8998 (ALL_SIMD): Add fp16fml.
8999 (armv8.2-a): Add fp16fml as an option.
9000 (armv8.3-a): Likewise.
9001 (armv8.4-a): Add fp16fml as part of fp16.
9002 * config/arm/arm.h (TARGET_FP16FML): Define.
9003 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9005 * config/arm/arm-modes.def (V2HF): Define.
9006 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9007 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9008 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9009 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9010 vfmsl_low, vfmsl_high): New set of builtins.
9011 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9012 (vfml_op): New code attribute.
9013 (VFMLHALVES): New int iterator.
9014 (VFML, VFMLSEL): New mode attributes.
9015 (V_reg): Define mapping for V2HF.
9016 (V_hi, V_lo): New mode attributes.
9017 (VF_constraint): Likewise.
9018 (vfml_half, vfml_half_selector): New int attributes.
9019 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9021 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9022 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9024 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9025 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9026 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9027 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9029 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9030 Document new effective target and option set.
9032 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9034 * config/arm/arm-cpus.in (armv8_4): New feature.
9035 (ARMv8_4a): New fgroup.
9036 (armv8.4-a): New arch.
9037 * config/arm/arm-tables.opt: Regenerate.
9038 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9039 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9040 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9041 Add matching rules for -march=armv8.4-a and extensions.
9042 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9044 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9047 * config/rx/rx.md (BW): New mode attribute.
9048 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9050 2018-01-11 Richard Biener <rguenther@suse.de>
9052 PR tree-optimization/83435
9053 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9054 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9055 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9057 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9058 Alan Hayward <alan.hayward@arm.com>
9059 David Sherwood <david.sherwood@arm.com>
9061 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9063 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9064 (aarch64_print_address_internal): Use it to check for a zero offset.
9066 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9067 Alan Hayward <alan.hayward@arm.com>
9068 David Sherwood <david.sherwood@arm.com>
9070 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9071 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9072 Return a poly_int64 rather than a HOST_WIDE_INT.
9073 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9074 rather than a HOST_WIDE_INT.
9075 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9076 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9077 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9078 final_offset from HOST_WIDE_INT to poly_int64.
9079 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9080 to_constant when getting the number of units in an Advanced SIMD
9082 (aarch64_builtin_vectorized_function): Check for a constant number
9084 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9086 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9087 attribute instead of GET_MODE_NUNITS.
9088 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9089 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9090 GET_MODE_SIZE for fixed-size registers.
9091 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9092 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9093 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9094 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9095 (aarch64_print_operand, aarch64_print_address_internal)
9096 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9097 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9098 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9099 Handle polynomial GET_MODE_SIZE.
9100 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9101 wider than SImode without modification.
9102 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9103 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9104 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9105 passing and returning SVE modes.
9106 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9107 rather than GEN_INT.
9108 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9109 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9110 (aarch64_allocate_and_probe_stack_space): Likewise.
9111 (aarch64_layout_frame): Cope with polynomial offsets.
9112 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9113 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9115 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9116 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9117 poly_int64 rather than a HOST_WIDE_INT.
9118 (aarch64_get_separate_components, aarch64_process_components)
9119 (aarch64_expand_prologue, aarch64_expand_epilogue)
9120 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9121 (aarch64_anchor_offset): New function, split out from...
9122 (aarch64_legitimize_address): ...here.
9123 (aarch64_builtin_vectorization_cost): Handle polynomial
9124 TYPE_VECTOR_SUBPARTS.
9125 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9127 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9128 number of elements from the PARALLEL rather than the mode.
9129 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9130 rather than GET_MODE_BITSIZE.
9131 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9132 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9133 (aarch64_expand_vec_perm_const_1): Handle polynomial
9134 d->perm.length () and d->perm elements.
9135 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9136 Apply to_constant to d->perm elements.
9137 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9138 polynomial CONST_VECTOR_NUNITS.
9139 (aarch64_move_pointer): Take amount as a poly_int64 rather
9141 (aarch64_progress_pointer): Avoid temporary variable.
9142 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9143 the mode attribute instead of GET_MODE.
9145 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9146 Alan Hayward <alan.hayward@arm.com>
9147 David Sherwood <david.sherwood@arm.com>
9149 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9150 x exists before using it.
9151 (aarch64_add_constant_internal): Rename to...
9152 (aarch64_add_offset_1): ...this. Replace regnum with separate
9153 src and dest rtxes. Handle the case in which they're different,
9154 including when the offset is zero. Replace scratchreg with an rtx.
9155 Use 2 additions if there is no spare register into which we can
9156 move a 16-bit constant.
9157 (aarch64_add_constant): Delete.
9158 (aarch64_add_offset): Replace reg with separate src and dest
9159 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9160 Use aarch64_add_offset_1.
9161 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9162 an rtx rather than an int. Take the delta as a poly_int64
9163 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9164 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9165 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9166 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9167 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9169 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9170 aarch64_add_constant.
9172 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9174 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9175 Use scalar_float_mode.
9177 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9179 * config/aarch64/aarch64-simd.md
9180 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9181 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9182 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9183 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9184 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9185 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9186 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9187 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9188 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9189 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9191 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9194 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9195 targ_options->x_arm_arch_string is non NULL.
9197 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9199 * config/aarch64/aarch64.h
9200 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9202 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9205 * expmed.c (emit_store_flag_force): Swap if const op0
9206 and change VOIDmode to mode of op0.
9208 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9210 PR rtl-optimization/83761
9211 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9212 than bytes to mode_for_size.
9214 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9217 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9218 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9221 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9224 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9225 when in layout mode.
9226 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9227 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9230 2018-01-10 Michael Collison <michael.collison@arm.com>
9232 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9233 * config/aarch64/aarch64-option-extension.def: Add
9234 AARCH64_OPT_EXTENSION of 'fp16fml'.
9235 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9236 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9237 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9238 * config/aarch64/constraints.md (Ui7): New constraint.
9239 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9240 (VFMLA_SEL_W): Ditto.
9243 (VFMLA16_LOW): New int iterator.
9244 (VFMLA16_HIGH): Ditto.
9245 (UNSPEC_FMLAL): New unspec.
9246 (UNSPEC_FMLSL): Ditto.
9247 (UNSPEC_FMLAL2): Ditto.
9248 (UNSPEC_FMLSL2): Ditto.
9249 (f16mac): New code attribute.
9250 * config/aarch64/aarch64-simd-builtins.def
9251 (aarch64_fmlal_lowv2sf): Ditto.
9252 (aarch64_fmlsl_lowv2sf): Ditto.
9253 (aarch64_fmlalq_lowv4sf): Ditto.
9254 (aarch64_fmlslq_lowv4sf): Ditto.
9255 (aarch64_fmlal_highv2sf): Ditto.
9256 (aarch64_fmlsl_highv2sf): Ditto.
9257 (aarch64_fmlalq_highv4sf): Ditto.
9258 (aarch64_fmlslq_highv4sf): Ditto.
9259 (aarch64_fmlal_lane_lowv2sf): Ditto.
9260 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9261 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9262 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9263 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9264 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9265 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9266 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9267 (aarch64_fmlal_lane_highv2sf): Ditto.
9268 (aarch64_fmlsl_lane_highv2sf): Ditto.
9269 (aarch64_fmlal_laneq_highv2sf): Ditto.
9270 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9271 (aarch64_fmlalq_lane_highv4sf): Ditto.
9272 (aarch64_fmlsl_lane_highv4sf): Ditto.
9273 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9274 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9275 * config/aarch64/aarch64-simd.md:
9276 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9277 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9278 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9279 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9280 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9281 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9282 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9283 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9284 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9285 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9286 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9287 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9288 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9289 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9290 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9291 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9292 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9293 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9294 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9295 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9296 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9297 (vfmlsl_low_u32): Ditto.
9298 (vfmlalq_low_u32): Ditto.
9299 (vfmlslq_low_u32): Ditto.
9300 (vfmlal_high_u32): Ditto.
9301 (vfmlsl_high_u32): Ditto.
9302 (vfmlalq_high_u32): Ditto.
9303 (vfmlslq_high_u32): Ditto.
9304 (vfmlal_lane_low_u32): Ditto.
9305 (vfmlsl_lane_low_u32): Ditto.
9306 (vfmlal_laneq_low_u32): Ditto.
9307 (vfmlsl_laneq_low_u32): Ditto.
9308 (vfmlalq_lane_low_u32): Ditto.
9309 (vfmlslq_lane_low_u32): Ditto.
9310 (vfmlalq_laneq_low_u32): Ditto.
9311 (vfmlslq_laneq_low_u32): Ditto.
9312 (vfmlal_lane_high_u32): Ditto.
9313 (vfmlsl_lane_high_u32): Ditto.
9314 (vfmlal_laneq_high_u32): Ditto.
9315 (vfmlsl_laneq_high_u32): Ditto.
9316 (vfmlalq_lane_high_u32): Ditto.
9317 (vfmlslq_lane_high_u32): Ditto.
9318 (vfmlalq_laneq_high_u32): Ditto.
9319 (vfmlslq_laneq_high_u32): Ditto.
9320 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9321 (AARCH64_FL_FOR_ARCH8_4): New.
9322 (AARCH64_ISA_F16FML): New ISA flag.
9323 (TARGET_F16FML): New feature flag for fp16fml.
9324 (doc/invoke.texi): Document new fp16fml option.
9326 2018-01-10 Michael Collison <michael.collison@arm.com>
9328 * config/aarch64/aarch64-builtins.c:
9329 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9330 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9331 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9332 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9333 (AARCH64_ISA_SHA3): New ISA flag.
9334 (TARGET_SHA3): New feature flag for sha3.
9335 * config/aarch64/iterators.md (sha512_op): New int attribute.
9336 (CRYPTO_SHA512): New int iterator.
9337 (UNSPEC_SHA512H): New unspec.
9338 (UNSPEC_SHA512H2): Ditto.
9339 (UNSPEC_SHA512SU0): Ditto.
9340 (UNSPEC_SHA512SU1): Ditto.
9341 * config/aarch64/aarch64-simd-builtins.def
9342 (aarch64_crypto_sha512hqv2di): New builtin.
9343 (aarch64_crypto_sha512h2qv2di): Ditto.
9344 (aarch64_crypto_sha512su0qv2di): Ditto.
9345 (aarch64_crypto_sha512su1qv2di): Ditto.
9346 (aarch64_eor3qv8hi): Ditto.
9347 (aarch64_rax1qv2di): Ditto.
9348 (aarch64_xarqv2di): Ditto.
9349 (aarch64_bcaxqv8hi): Ditto.
9350 * config/aarch64/aarch64-simd.md:
9351 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9352 (aarch64_crypto_sha512su0qv2di): Ditto.
9353 (aarch64_crypto_sha512su1qv2di): Ditto.
9354 (aarch64_eor3qv8hi): Ditto.
9355 (aarch64_rax1qv2di): Ditto.
9356 (aarch64_xarqv2di): Ditto.
9357 (aarch64_bcaxqv8hi): Ditto.
9358 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9359 (vsha512h2q_u64): Ditto.
9360 (vsha512su0q_u64): Ditto.
9361 (vsha512su1q_u64): Ditto.
9362 (veor3q_u16): Ditto.
9363 (vrax1q_u64): Ditto.
9365 (vbcaxq_u16): Ditto.
9366 * config/arm/types.md (crypto_sha512): New type attribute.
9367 (crypto_sha3): Ditto.
9368 (doc/invoke.texi): Document new sha3 option.
9370 2018-01-10 Michael Collison <michael.collison@arm.com>
9372 * config/aarch64/aarch64-builtins.c:
9373 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9374 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9375 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9376 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9377 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9378 (AARCH64_ISA_SM4): New ISA flag.
9379 (TARGET_SM4): New feature flag for sm4.
9380 * config/aarch64/aarch64-simd-builtins.def
9381 (aarch64_sm3ss1qv4si): Ditto.
9382 (aarch64_sm3tt1aq4si): Ditto.
9383 (aarch64_sm3tt1bq4si): Ditto.
9384 (aarch64_sm3tt2aq4si): Ditto.
9385 (aarch64_sm3tt2bq4si): Ditto.
9386 (aarch64_sm3partw1qv4si): Ditto.
9387 (aarch64_sm3partw2qv4si): Ditto.
9388 (aarch64_sm4eqv4si): Ditto.
9389 (aarch64_sm4ekeyqv4si): Ditto.
9390 * config/aarch64/aarch64-simd.md:
9391 (aarch64_sm3ss1qv4si): Ditto.
9392 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9393 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9394 (aarch64_sm4eqv4si): Ditto.
9395 (aarch64_sm4ekeyqv4si): Ditto.
9396 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9397 (sm3part_op): Ditto.
9398 (CRYPTO_SM3TT): Ditto.
9399 (CRYPTO_SM3PART): Ditto.
9400 (UNSPEC_SM3SS1): New unspec.
9401 (UNSPEC_SM3TT1A): Ditto.
9402 (UNSPEC_SM3TT1B): Ditto.
9403 (UNSPEC_SM3TT2A): Ditto.
9404 (UNSPEC_SM3TT2B): Ditto.
9405 (UNSPEC_SM3PARTW1): Ditto.
9406 (UNSPEC_SM3PARTW2): Ditto.
9407 (UNSPEC_SM4E): Ditto.
9408 (UNSPEC_SM4EKEY): Ditto.
9409 * config/aarch64/constraints.md (Ui2): New constraint.
9410 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9411 * config/arm/types.md (crypto_sm3): New type attribute.
9412 (crypto_sm4): Ditto.
9413 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9414 (vsm3tt1aq_u32): Ditto.
9415 (vsm3tt1bq_u32): Ditto.
9416 (vsm3tt2aq_u32): Ditto.
9417 (vsm3tt2bq_u32): Ditto.
9418 (vsm3partw1q_u32): Ditto.
9419 (vsm3partw2q_u32): Ditto.
9420 (vsm4eq_u32): Ditto.
9421 (vsm4ekeyq_u32): Ditto.
9422 (doc/invoke.texi): Document new sm4 option.
9424 2018-01-10 Michael Collison <michael.collison@arm.com>
9426 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9427 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9428 (AARCH64_FL_FOR_ARCH8_4): New.
9429 (AARCH64_FL_V8_4): New flag.
9430 (doc/invoke.texi): Document new armv8.4-a option.
9432 2018-01-10 Michael Collison <michael.collison@arm.com>
9434 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9435 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9436 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9437 * config/aarch64/aarch64-option-extension.def: Add
9438 AARCH64_OPT_EXTENSION of 'sha2'.
9439 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9440 (crypto): Disable sha2 and aes if crypto disabled.
9441 (crypto): Enable aes and sha2 if enabled.
9442 (simd): Disable sha2 and aes if simd disabled.
9443 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9445 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9446 (TARGET_SHA2): New feature flag for sha2.
9447 (TARGET_AES): New feature flag for aes.
9448 * config/aarch64/aarch64-simd.md:
9449 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9450 conditional on TARGET_AES.
9451 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9452 (aarch64_crypto_sha1hsi): Make pattern conditional
9454 (aarch64_crypto_sha1hv4si): Ditto.
9455 (aarch64_be_crypto_sha1hv4si): Ditto.
9456 (aarch64_crypto_sha1su1v4si): Ditto.
9457 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9458 (aarch64_crypto_sha1su0v4si): Ditto.
9459 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9460 (aarch64_crypto_sha256su0v4si): Ditto.
9461 (aarch64_crypto_sha256su1v4si): Ditto.
9462 (doc/invoke.texi): Document new aes and sha2 options.
9464 2018-01-10 Martin Sebor <msebor@redhat.com>
9466 PR tree-optimization/83781
9467 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9470 2018-01-11 Martin Sebor <msebor@gmail.com>
9471 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9473 PR tree-optimization/83501
9474 PR tree-optimization/81703
9476 * tree-ssa-strlen.c (get_string_cst): Rename...
9477 (get_string_len): ...to this. Handle global constants.
9478 (handle_char_store): Adjust.
9480 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9481 Jim Wilson <jimw@sifive.com>
9483 * config/riscv/riscv-protos.h (riscv_output_return): New.
9484 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9485 (riscv_attribute_table, riscv_output_return),
9486 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9487 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9488 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9490 (riscv_expand_prologue): Add early return for naked function.
9491 (riscv_expand_epilogue): Likewise.
9492 (riscv_function_ok_for_sibcall): Return false for naked function.
9493 (riscv_set_current_function): New.
9494 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9495 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9496 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9497 * doc/extend.texi (RISC-V Function Attributes): New.
9499 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9501 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9502 check for 128-bit long double before checking TCmode.
9503 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9504 128-bit long doubles before checking TFmode or TCmode.
9505 (FLOAT128_IBM_P): Likewise.
9507 2018-01-10 Martin Sebor <msebor@redhat.com>
9509 PR tree-optimization/83671
9510 * builtins.c (c_strlen): Unconditionally return zero for the empty
9512 Use -Warray-bounds for warnings.
9513 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9514 for non-constant array indices with COMPONENT_REF, arrays of
9515 arrays, and pointers to arrays.
9516 (gimple_fold_builtin_strlen): Determine and set length range for
9517 non-constant character arrays.
9519 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9522 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9525 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9527 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9529 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9532 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9533 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9534 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9535 indexed_or_indirect_operand predicate.
9536 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9537 (*vsx_le_perm_load_v8hi): Likewise.
9538 (*vsx_le_perm_load_v16qi): Likewise.
9539 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9540 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9541 (*vsx_le_perm_store_v8hi): Likewise.
9542 (*vsx_le_perm_store_v16qi): Likewise.
9543 (eight unnamed splitters): Likewise.
9545 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9547 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9548 * config/rs6000/emmintrin.h: Likewise.
9549 * config/rs6000/mmintrin.h: Likewise.
9550 * config/rs6000/xmmintrin.h: Likewise.
9552 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9555 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9557 * tree.c (tree_nop_conversion): Return true for location wrapper
9559 (maybe_wrap_with_location): New function.
9560 (selftest::check_strip_nops): New function.
9561 (selftest::test_location_wrappers): New function.
9562 (selftest::tree_c_tests): Call it.
9563 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9564 (maybe_wrap_with_location): New decl.
9565 (EXPR_LOCATION_WRAPPER_P): New macro.
9566 (location_wrapper_p): New inline function.
9567 (tree_strip_any_location_wrapper): New inline function.
9569 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9572 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9573 stack_realign_offset for the largest alignment of stack slot
9575 (ix86_find_max_used_stack_alignment): New function.
9576 (ix86_finalize_stack_frame_flags): Use it. Set
9577 max_used_stack_alignment if we don't realign stack.
9578 * config/i386/i386.h (machine_function): Add
9579 max_used_stack_alignment.
9581 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9583 * config/arm/arm.opt (-mbranch-cost): New option.
9584 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9587 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9590 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9591 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9593 2018-01-10 Richard Biener <rguenther@suse.de>
9596 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9597 early out so it also covers the case where we have a non-NULL
9600 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9602 PR tree-optimization/83753
9603 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9604 for non-strided grouped accesses if the number of elements is 1.
9606 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9609 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9610 * i386.h (TARGET_USE_GATHER): Define.
9611 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9613 2018-01-10 Martin Liska <mliska@suse.cz>
9616 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9617 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9619 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9620 CLEANUP_NO_PARTITIONING is not set.
9622 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9624 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9625 for vectors, as a partial revert of r254296.
9626 * rtl.h (const_vec_p): Delete.
9627 (const_vec_duplicate_p): Don't test for vector CONSTs.
9628 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9629 * expmed.c (make_tree): Likewise.
9632 * common.md (E, F): Use CONSTANT_P instead of checking for
9634 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9635 checking for CONST_VECTOR.
9637 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9640 * predict.c (force_edge_cold): Handle in more sane way edges
9643 2018-01-09 Carl Love <cel@us.ibm.com>
9645 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9647 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9648 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9649 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9650 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9651 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9652 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9653 * config/rs6000/rs6000-protos.h: Add extern defition for
9654 rs6000_generate_float2_double_code.
9655 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9657 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9658 (float2_v2df): Add define_expand.
9660 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9663 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9664 op_mode in the force_to_mode call.
9666 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9668 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9669 instead of checking each element individually.
9670 (aarch64_evpc_uzp): Likewise.
9671 (aarch64_evpc_zip): Likewise.
9672 (aarch64_evpc_ext): Likewise.
9673 (aarch64_evpc_rev): Likewise.
9674 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9675 instead of checking each element individually. Return true without
9677 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9678 whether all selected elements come from the same input, instead of
9679 checking each element individually. Remove calls to gen_rtx_REG,
9680 start_sequence and end_sequence and instead assert that no rtl is
9683 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9685 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9686 order of HIGH and CONST checks.
9688 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9690 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9691 if the destination isn't an SSA_NAME.
9693 2018-01-09 Richard Biener <rguenther@suse.de>
9695 PR tree-optimization/83668
9696 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9698 (canonicalize_loop_form): ... here, renamed from ...
9699 (canonicalize_loop_closed_ssa_form): ... this and amended to
9700 swap successor edges for loop exit blocks to make us use
9701 the RPO order we need for initial schedule generation.
9703 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9705 PR tree-optimization/64811
9706 * match.pd: When optimizing comparisons with Inf, avoid
9707 introducing or losing exceptions from comparisons with NaN.
9709 2018-01-09 Martin Liska <mliska@suse.cz>
9712 * asan.c (shadow_mem_size): Add gcc_assert.
9714 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9716 Don't save registers in main().
9719 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9720 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9721 * config/avr/avr.c (avr_set_current_function): Don't error if
9722 naked, OS_task or OS_main are specified at the same time.
9723 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9725 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9727 * common/config/avr/avr-common.c (avr_option_optimization_table):
9728 Switch on -mmain-is-OS_task for optimizing compilations.
9730 2018-01-09 Richard Biener <rguenther@suse.de>
9732 PR tree-optimization/83572
9733 * graphite.c: Include cfganal.h.
9734 (graphite_transform_loops): Connect infinite loops to exit
9735 and remove fake edges at the end.
9737 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9739 * ipa-inline.c (edge_badness): Revert accidental checkin.
9741 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9744 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9745 symbols; not inline clones.
9747 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9750 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9751 hard registers. Formatting fixes.
9753 PR preprocessor/83722
9754 * gcc.c (try_generate_repro): Pass
9755 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9756 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9759 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9760 Kito Cheng <kito.cheng@gmail.com>
9762 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9763 (riscv_leaf_function_p): Delete.
9764 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9766 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9768 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9770 (do_ifelse): New function.
9771 (do_isel): New function.
9772 (do_sub3): New function.
9773 (do_add3): New function.
9774 (do_load_mask_compare): New function.
9775 (do_overlap_load_compare): New function.
9776 (expand_compare_loop): New function.
9777 (expand_block_compare): Call expand_compare_loop() when appropriate.
9778 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9780 (-mblock-compare-inline-loop-limit): New option.
9782 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9785 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9786 Reverse order of second and third operands in first alternative.
9787 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9788 of first and second elements in UNSPEC_VPERMR vector.
9789 (altivec_expand_vec_perm_le): Likewise.
9791 2018-01-08 Jeff Law <law@redhat.com>
9793 PR rtl-optimizatin/81308
9794 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9795 (process_switch): If group_case_labels makes a change, then set
9797 (pass_convert_switch::execute): If a switch is converted, then
9798 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9800 PR rtl-optimization/81308
9801 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9804 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9806 PR target/83663 - Revert r255946
9807 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9808 generation for cases where splatting a value is not useful.
9809 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9810 across a vec_duplicate and a paradoxical subreg forming a vector
9811 mode to a vec_concat.
9813 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9815 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9816 -march=armv8.3-a variants.
9817 * config/arm/t-multilib: Likewise.
9818 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9820 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9822 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9824 (cceq_ior_compare_complement): Give it a name so I can use it, and
9825 change boolean_or_operator predicate to boolean_operator so it can
9826 be used to generate a crand.
9827 (eqne): New code iterator.
9828 (bd/bd_neg): New code_attrs.
9829 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9830 a single define_insn.
9831 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9832 decrement (bdnzt/bdnzf/bdzt/bdzf).
9833 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9834 with the new names of the branch decrement patterns, and added the
9835 names of the branch decrement conditional patterns.
9837 2018-01-08 Richard Biener <rguenther@suse.de>
9839 PR tree-optimization/83563
9840 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9843 2018-01-08 Richard Biener <rguenther@suse.de>
9846 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9848 2018-01-08 Richard Biener <rguenther@suse.de>
9850 PR tree-optimization/83685
9851 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9852 references to abnormals.
9854 2018-01-08 Richard Biener <rguenther@suse.de>
9857 * dwarf2out.c (output_indirect_strings): Handle empty
9858 skeleton_debug_str_hash.
9859 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9861 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9863 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9864 (emit_store_direct): Likewise.
9865 (arc_trampoline_adjust_address): Likewise.
9866 (arc_asm_trampoline_template): New function.
9867 (arc_initialize_trampoline): Use asm_trampoline_template.
9868 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9869 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9870 * config/arc/arc.md (flush_icache): Delete pattern.
9872 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9874 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9875 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9878 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9881 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9882 by not USED_FOR_TARGET.
9883 (make_pass_resolve_sw_modes): Likewise.
9885 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9887 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9890 2018-01-08 Richard Biener <rguenther@suse.de>
9893 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9895 2018-01-08 Richard Biener <rguenther@suse.de>
9898 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9900 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9903 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9904 basic blocks with a small number of successors.
9905 (convert_control_dep_chain_into_preds): Improve handling of
9907 (dump_predicates): Split apart into...
9908 (dump_pred_chain): ...here...
9909 (dump_pred_info): ...and here.
9910 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9911 (can_chain_union_be_invalidated_p): Improve check for invalidation
9913 (uninit_uses_cannot_happen): Avoid unnecessary if
9914 convert_control_dep_chain_into_preds yielded nothing.
9916 2018-01-06 Martin Sebor <msebor@redhat.com>
9918 PR tree-optimization/83640
9919 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9920 subtracting negative offset from size.
9921 (builtin_access::overlap): Adjust offset bounds of the access to fall
9922 within the size of the object if possible.
9924 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9926 PR rtl-optimization/83699
9927 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9928 extract_bit_field_as_subreg to cases in which the extracted
9929 value is also a vector.
9931 * lra-constraints.c (process_alt_operands): Test for the equivalence
9932 substitutions when detecting a possible reload cycle.
9934 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9937 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9938 by default if flag_selective_schedling{,2}. Formatting fixes.
9940 PR rtl-optimization/83682
9941 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9942 if it has non-VECTOR_MODE element mode.
9943 (vec_duplicate_p): Likewise.
9946 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9947 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9949 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9952 * config/i386/i386-builtin.def
9953 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9954 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9955 Require also OPTION_MASK_ISA_AVX512F in addition to
9956 OPTION_MASK_ISA_GFNI.
9957 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9958 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9959 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9960 to OPTION_MASK_ISA_GFNI.
9961 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9962 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9963 OPTION_MASK_ISA_AVX512BW.
9964 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9965 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9966 addition to OPTION_MASK_ISA_GFNI.
9967 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9968 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9969 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9970 to OPTION_MASK_ISA_GFNI.
9971 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9972 a requirement for all ISAs rather than any of them with a few
9974 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9976 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9977 bitmasks to be enabled with 3 exceptions, instead of requiring any
9978 enabled ISA with lots of exceptions.
9979 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9980 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9981 Change avx512bw in isa attribute to avx512f.
9982 * config/i386/sgxintrin.h: Add license boilerplate.
9983 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9984 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9985 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9986 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9988 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9989 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9990 temporarily sse2 rather than sse if not enabled already.
9993 * config/i386/sse.md (VI248_VLBW): Rename to ...
9994 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9995 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9996 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9997 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9998 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9999 mode iterator instead of VI248_VLBW.
10001 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10003 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10004 (record_modified): Skip clobbers; add debug output.
10005 (param_change_prob): Use sreal frequencies.
10007 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10009 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10010 punt for user-aligned variables.
10012 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10014 * tree-chrec.c (chrec_contains_symbols): Return true for
10017 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10020 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10021 of (x|y) == x for BICS pattern.
10023 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10025 PR tree-optimization/83605
10026 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10027 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10030 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10032 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10033 * config/epiphany/rtems.h: New file.
10035 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10036 Uros Bizjak <ubizjak@gmail.com>
10039 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10040 QIreg_operand instead of register_operand predicate.
10041 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10042 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10043 comments instead of -fmitigate[-_]rop.
10045 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10048 * cgraphunit.c (symbol_table::compile): Switch to text_section
10049 before calling assembly_start debug hook.
10050 * run-rtl-passes.c (run_rtl_passes): Likewise.
10053 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10055 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10056 range_int_cst_p rather than !symbolic_range_p before calling
10057 extract_range_from_multiplicative_op_1.
10059 2018-01-04 Jeff Law <law@redhat.com>
10061 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10062 redundant test in assertion.
10064 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10066 * doc/rtl.texi: Document machine_mode wrapper classes.
10068 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10070 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10071 using tree_to_uhwi.
10073 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10075 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10076 the VEC_PERM_EXPR fold to fail.
10078 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10081 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10082 to switched_sections.
10084 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10087 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10088 test for d.testing.
10090 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10093 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10094 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10096 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10099 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10100 is BLKmode and bitpos not zero or mode change is needed.
10102 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10105 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10108 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10111 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10112 instead of MULT rtx. Update all corresponding splitters.
10113 (*saddl_se): Ditto.
10114 (*ssub<modesuffix>): Ditto.
10115 (*ssubl_se): Ditto.
10116 (*cmp_sadd_di): Update split patterns.
10117 (*cmp_sadd_si): Ditto.
10118 (*cmp_sadd_sidi): Ditto.
10119 (*cmp_ssub_di): Ditto.
10120 (*cmp_ssub_si): Ditto.
10121 (*cmp_ssub_sidi): Ditto.
10122 * config/alpha/predicates.md (const23_operand): New predicate.
10123 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10124 Look for ASHIFT, not MULT inner operand.
10125 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10127 2018-01-04 Martin Liska <mliska@suse.cz>
10129 PR gcov-profile/83669
10130 * gcov.c (output_intermediate_file): Add version to intermediate
10132 * doc/gcov.texi: Document new field 'version' in intermediate
10133 file format. Fix location of '-k' option of gcov command.
10135 2018-01-04 Martin Liska <mliska@suse.cz>
10138 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10140 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10142 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10144 2018-01-03 Martin Sebor <msebor@redhat.com>
10146 PR tree-optimization/83655
10147 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10148 checking calls with invalid arguments.
10150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10152 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10153 (vectorizable_mask_load_store): Delete.
10154 (vectorizable_call): Return false for masked loads and stores.
10155 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10156 instead of gimple_assign_rhs1.
10157 (vectorizable_load): Handle IFN_MASK_LOAD.
10158 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10162 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10164 (vectorizable_mask_load_store): ...here.
10165 (vectorizable_load): ...and here.
10167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10169 * tree-vect-stmts.c (vect_build_all_ones_mask)
10170 (vect_build_zero_merge_argument): New functions, split out from...
10171 (vectorizable_load): ...here.
10173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10175 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10177 (vectorizable_mask_load_store): ...here.
10178 (vectorizable_store): ...and here.
10180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10182 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10184 (vectorizable_mask_load_store): ...here.
10186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10188 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10189 (vect_model_store_cost): Take a vec_load_store_type instead of a
10191 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10192 (vect_model_store_cost): Take a vec_load_store_type instead of a
10194 (vectorizable_mask_load_store): Update accordingly.
10195 (vectorizable_store): Likewise.
10196 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10200 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10201 IFN_MASK_LOAD calls here rather than...
10202 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10205 Alan Hayward <alan.hayward@arm.com>
10206 David Sherwood <david.sherwood@arm.com>
10208 * expmed.c (extract_bit_field_1): For vector extracts,
10209 fall back to extract_bit_field_as_subreg if vec_extract
10212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10213 Alan Hayward <alan.hayward@arm.com>
10214 David Sherwood <david.sherwood@arm.com>
10216 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10217 they are variable or constant sized.
10218 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10219 slots for constant-sized data.
10221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10222 Alan Hayward <alan.hayward@arm.com>
10223 David Sherwood <david.sherwood@arm.com>
10225 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10226 handling COND_EXPRs with boolean comparisons, try to find a better
10227 basis for the mask type than the boolean itself.
10229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10231 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10232 is calculated and how it can be overridden.
10233 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10234 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10236 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10239 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10240 Alan Hayward <alan.hayward@arm.com>
10241 David Sherwood <david.sherwood@arm.com>
10243 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10244 Remove the mode argument.
10245 (aarch64_simd_valid_immediate): Remove the mode and inverse
10247 * config/aarch64/iterators.md (bitsize): New iterator.
10248 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10249 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10250 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10251 aarch64_simd_valid_immediate.
10252 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10253 (aarch64_reg_or_bic_imm): Likewise.
10254 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10255 with an insn_type enum and msl with a modifier_type enum.
10256 Replace element_width with a scalar_mode. Change the shift
10257 to unsigned int. Add constructors for scalar_float_mode and
10258 scalar_int_mode elements.
10259 (aarch64_vect_float_const_representable_p): Delete.
10260 (aarch64_can_const_movi_rtx_p)
10261 (aarch64_simd_scalar_immediate_valid_for_move)
10262 (aarch64_simd_make_constant): Update call to
10263 aarch64_simd_valid_immediate.
10264 (aarch64_advsimd_valid_immediate_hs): New function.
10265 (aarch64_advsimd_valid_immediate): Likewise.
10266 (aarch64_simd_valid_immediate): Remove mode and inverse
10267 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10268 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10269 and aarch64_float_const_representable_p on the result.
10270 (aarch64_output_simd_mov_immediate): Remove mode argument.
10271 Update call to aarch64_simd_valid_immediate and use of
10272 simd_immediate_info.
10273 (aarch64_output_scalar_simd_mov_immediate): Update call
10276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10277 Alan Hayward <alan.hayward@arm.com>
10278 David Sherwood <david.sherwood@arm.com>
10280 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10281 (mode_nunits): Likewise CONST_MODE_NUNITS.
10282 * machmode.def (ADJUST_NUNITS): Document.
10283 * genmodes.c (mode_data::need_nunits_adj): New field.
10284 (blank_mode): Update accordingly.
10285 (adj_nunits): New variable.
10286 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10288 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10289 listed in adj_nunits.
10290 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10291 listed in adj_nunits. Don't emit case statements for such modes.
10292 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10293 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10294 nothing if adj_nunits is nonnull.
10295 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10296 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10297 (emit_mode_fbit): Update use of print_maybe_const_decl.
10298 (emit_move_size): Likewise. Treat the array as non-const
10300 (emit_mode_adjustments): Handle adj_nunits.
10302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10304 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10305 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10306 (VECTOR_MODES): Use it.
10307 (make_vector_modes): Take the prefix as an argument.
10309 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10310 Alan Hayward <alan.hayward@arm.com>
10311 David Sherwood <david.sherwood@arm.com>
10313 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10314 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10315 for MODE_VECTOR_BOOL.
10316 * machmode.def (VECTOR_BOOL_MODE): Document.
10317 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10318 (make_vector_bool_mode): New function.
10319 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10321 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10322 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10324 * stor-layout.c (int_mode_for_mode): Likewise.
10325 * tree.c (build_vector_type_for_mode): Likewise.
10326 * varasm.c (output_constant_pool_2): Likewise.
10327 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10328 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10329 for MODE_VECTOR_BOOL.
10330 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10331 of mode class checks.
10332 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10333 instead of a list of mode class checks.
10334 (expand_vector_scalar_condition): Likewise.
10335 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10338 Alan Hayward <alan.hayward@arm.com>
10339 David Sherwood <david.sherwood@arm.com>
10341 * machmode.h (mode_size): Change from unsigned short to
10343 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10344 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10345 or if measurement_type is not polynomial.
10346 (fixed_size_mode::includes_p): Check for constant-sized modes.
10347 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10348 return a poly_uint16 rather than an unsigned short.
10349 (emit_mode_size): Change the type of mode_size from unsigned short
10350 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10351 (emit_mode_adjustments): Cope with polynomial vector sizes.
10352 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10354 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10356 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10357 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10358 * caller-save.c (setup_save_areas): Likewise.
10359 (replace_reg_with_saved_mem): Likewise.
10360 * calls.c (emit_library_call_value_1): Likewise.
10361 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10362 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10363 (gen_lowpart_for_combine): Likewise.
10364 * convert.c (convert_to_integer_1): Likewise.
10365 * cse.c (equiv_constant, cse_insn): Likewise.
10366 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10367 (cselib_subst_to_values): Likewise.
10368 * dce.c (word_dce_process_block): Likewise.
10369 * df-problems.c (df_word_lr_mark_ref): Likewise.
10370 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10371 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10372 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10373 (rtl_for_decl_location): Likewise.
10374 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10375 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10376 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10377 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10378 (expand_expr_real_1): Likewise.
10379 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10380 (pad_below): Likewise.
10381 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10382 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10383 * ira.c (get_subreg_tracking_sizes): Likewise.
10384 * ira-build.c (ira_create_allocno_objects): Likewise.
10385 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10386 (ira_sort_regnos_for_alter_reg): Likewise.
10387 * ira-costs.c (record_operand_costs): Likewise.
10388 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10389 (resolve_simple_move): Likewise.
10390 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10391 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10392 (lra_constraints): Likewise.
10393 (CONST_POOL_OK_P): Reject variable-sized modes.
10394 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10395 (add_pseudo_to_slot, lra_spill): Likewise.
10396 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10397 * optabs-query.c (get_best_extraction_insn): Likewise.
10398 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10399 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10400 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10401 * recog.c (offsettable_address_addr_space_p): Likewise.
10402 * regcprop.c (maybe_mode_change): Likewise.
10403 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10404 * regrename.c (build_def_use): Likewise.
10405 * regstat.c (dump_reg_info): Likewise.
10406 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10407 (find_reloads, find_reloads_subreg_address): Likewise.
10408 * reload1.c (eliminate_regs_1): Likewise.
10409 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10410 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10411 (simplify_binary_operation_1, simplify_subreg): Likewise.
10412 * targhooks.c (default_function_arg_padding): Likewise.
10413 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10414 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10415 (verify_gimple_assign_ternary): Likewise.
10416 * tree-inline.c (estimate_move_cost): Likewise.
10417 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10418 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10419 (get_address_cost_ainc): Likewise.
10420 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10421 (vect_supportable_dr_alignment): Likewise.
10422 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10423 (vectorizable_reduction): Likewise.
10424 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10425 (vectorizable_operation, vectorizable_load): Likewise.
10426 * tree.c (build_same_sized_truth_vector_type): Likewise.
10427 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10428 * var-tracking.c (emit_note_insn_var_location): Likewise.
10429 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10430 (ADDR_VEC_ALIGN): Likewise.
10432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10433 Alan Hayward <alan.hayward@arm.com>
10434 David Sherwood <david.sherwood@arm.com>
10436 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10438 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10439 or if measurement_type is polynomial.
10440 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10441 * combine.c (make_extraction): Likewise.
10442 * dse.c (find_shift_sequence): Likewise.
10443 * dwarf2out.c (mem_loc_descriptor): Likewise.
10444 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10445 (extract_bit_field, extract_low_bits): Likewise.
10446 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10447 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10448 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10449 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10450 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10451 * reload.c (find_reloads): Likewise.
10452 * reload1.c (alter_reg): Likewise.
10453 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10454 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10455 * tree-if-conv.c (predicate_mem_writes): Likewise.
10456 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10457 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10458 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10459 * valtrack.c (dead_debug_insert_temp): Likewise.
10460 * varasm.c (mergeable_constant_section): Likewise.
10461 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10464 Alan Hayward <alan.hayward@arm.com>
10465 David Sherwood <david.sherwood@arm.com>
10467 * expr.c (expand_assignment): Cope with polynomial mode sizes
10468 when assigning to a CONCAT.
10470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10471 Alan Hayward <alan.hayward@arm.com>
10472 David Sherwood <david.sherwood@arm.com>
10474 * machmode.h (mode_precision): Change from unsigned short to
10476 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10478 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10479 or if measurement_type is not polynomial.
10480 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10481 in which the mode is already known to be a scalar_int_mode.
10482 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10483 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10485 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10486 for GET_MODE_PRECISION.
10487 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10488 for GET_MODE_PRECISION.
10489 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10491 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10492 (expand_field_assignment, make_extraction): Likewise.
10493 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10494 (get_last_value): Likewise.
10495 * convert.c (convert_to_integer_1): Likewise.
10496 * cse.c (cse_insn): Likewise.
10497 * expr.c (expand_expr_real_1): Likewise.
10498 * lra-constraints.c (simplify_operand_subreg): Likewise.
10499 * optabs-query.c (can_atomic_load_p): Likewise.
10500 * optabs.c (expand_atomic_load): Likewise.
10501 (expand_atomic_store): Likewise.
10502 * ree.c (combine_reaching_defs): Likewise.
10503 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10504 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10505 * tree.h (type_has_mode_precision_p): Likewise.
10506 * ubsan.c (instrument_si_overflow): Likewise.
10508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10509 Alan Hayward <alan.hayward@arm.com>
10510 David Sherwood <david.sherwood@arm.com>
10512 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10513 polynomial numbers of units.
10514 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10515 (valid_vector_subparts_p): New function.
10516 (build_vector_type): Remove temporary shim and take the number
10517 of units as a poly_uint64 rather than an int.
10518 (build_opaque_vector_type): Take the number of units as a
10519 poly_uint64 rather than an int.
10520 * tree.c (build_vector_from_ctor): Handle polynomial
10521 TYPE_VECTOR_SUBPARTS.
10522 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10523 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10524 (build_vector_from_val): If the number of units is variable,
10525 use build_vec_duplicate_cst for constant operands and
10526 VEC_DUPLICATE_EXPR otherwise.
10527 (make_vector_type): Remove temporary is_constant ().
10528 (build_vector_type, build_opaque_vector_type): Take the number of
10529 units as a poly_uint64 rather than an int.
10530 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10532 * cfgexpand.c (expand_debug_expr): Likewise.
10533 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10534 (store_constructor, expand_expr_real_1): Likewise.
10535 (const_scalar_mask_from_tree): Likewise.
10536 * fold-const-call.c (fold_const_reduction): Likewise.
10537 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10538 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10539 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10540 (fold_relational_const): Likewise.
10541 (native_interpret_vector): Likewise. Change the size from an
10542 int to an unsigned int.
10543 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10544 TYPE_VECTOR_SUBPARTS.
10545 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10546 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10547 duplicating a non-constant operand into a variable-length vector.
10548 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10549 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10550 * ipa-icf.c (sem_variable::equals): Likewise.
10551 * match.pd: Likewise.
10552 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10553 * print-tree.c (print_node): Likewise.
10554 * stor-layout.c (layout_type): Likewise.
10555 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10556 * tree-cfg.c (verify_gimple_comparison): Likewise.
10557 (verify_gimple_assign_binary): Likewise.
10558 (verify_gimple_assign_ternary): Likewise.
10559 (verify_gimple_assign_single): Likewise.
10560 * tree-pretty-print.c (dump_generic_node): Likewise.
10561 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10562 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10563 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10564 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10565 (vect_shift_permute_load_chain): Likewise.
10566 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10567 (expand_vector_condition, optimize_vector_constructor): Likewise.
10568 (lower_vec_perm, get_compute_type): Likewise.
10569 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10570 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10571 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10572 (vect_recog_mask_conversion_pattern): Likewise.
10573 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10574 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10575 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10576 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10577 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10578 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10579 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10580 (supportable_widening_operation): Likewise.
10581 (supportable_narrowing_operation): Likewise.
10582 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10584 * varasm.c (output_constant): Likewise.
10586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10587 Alan Hayward <alan.hayward@arm.com>
10588 David Sherwood <david.sherwood@arm.com>
10590 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10591 so that both the length == 3 and length != 3 cases set up their
10592 own permute vectors. Add comments explaining why we know the
10593 number of elements is constant.
10594 (vect_permute_load_chain): Likewise.
10596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10597 Alan Hayward <alan.hayward@arm.com>
10598 David Sherwood <david.sherwood@arm.com>
10600 * machmode.h (mode_nunits): Change from unsigned char to
10602 (ONLY_FIXED_SIZE_MODES): New macro.
10603 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10604 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10605 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10607 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10608 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10609 or if measurement_type is not polynomial.
10610 * genmodes.c (ZERO_COEFFS): New macro.
10611 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10613 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10614 Use ZERO_COEFFS when emitting initializers.
10615 * data-streamer.h (bp_pack_poly_value): New function.
10616 (bp_unpack_poly_value): Likewise.
10617 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10618 for GET_MODE_NUNITS.
10619 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10620 for GET_MODE_NUNITS.
10621 * tree.c (make_vector_type): Remove temporary shim and make
10622 the real function take the number of units as a poly_uint64
10623 rather than an int.
10624 (build_vector_type_for_mode): Handle polynomial nunits.
10625 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10626 * emit-rtl.c (const_vec_series_p_1): Likewise.
10627 (gen_rtx_CONST_VECTOR): Likewise.
10628 * fold-const.c (test_vec_duplicate_folding): Likewise.
10629 * genrecog.c (validate_pattern): Likewise.
10630 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10631 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10632 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10633 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10634 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10635 * rtlanal.c (subreg_get_info): Likewise.
10636 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10637 (vect_grouped_load_supported): Likewise.
10638 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10639 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10640 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10641 (simplify_const_unary_operation, simplify_binary_operation_1)
10642 (simplify_const_binary_operation, simplify_ternary_operation)
10643 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10644 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10645 instead of CONST_VECTOR_NUNITS.
10646 * varasm.c (output_constant_pool_2): Likewise.
10647 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10648 explicit-encoded elements in the XVEC for variable-length vectors.
10650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10652 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10655 Alan Hayward <alan.hayward@arm.com>
10656 David Sherwood <david.sherwood@arm.com>
10658 * coretypes.h (fixed_size_mode): Declare.
10659 (fixed_size_mode_pod): New typedef.
10660 * builtins.h (target_builtins::x_apply_args_mode)
10661 (target_builtins::x_apply_result_mode): Change type to
10662 fixed_size_mode_pod.
10663 * builtins.c (apply_args_size, apply_result_size, result_vector)
10664 (expand_builtin_apply_args_1, expand_builtin_apply)
10665 (expand_builtin_return): Update accordingly.
10667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10669 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10670 * cselib.c (cselib_hash_rtx): Likewise.
10671 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10672 CONST_VECTOR encoding.
10674 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10675 Jeff Law <law@redhat.com>
10678 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10679 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10680 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10681 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10684 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10685 explicitly probe *sp in a noreturn function if there were any callee
10686 register saves or frame pointer is needed.
10688 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10691 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10692 BLKmode for ternary, binary or unary expressions.
10695 * var-tracking.c (delete_vta_debug_insn): New inline function.
10696 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10697 insns from get_insns () to NULL instead of each bb separately.
10698 Use delete_vta_debug_insn. No longer static.
10699 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10700 delete_vta_debug_insns callers.
10701 * rtl.h (delete_vta_debug_insns): Declare.
10702 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10703 instead of variable_tracking_main.
10705 2018-01-03 Martin Sebor <msebor@redhat.com>
10707 PR tree-optimization/83603
10708 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10709 arguments past the endof the argument list in functions declared
10710 without a prototype.
10711 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10712 Avoid checking when arguments are null.
10714 2018-01-03 Martin Sebor <msebor@redhat.com>
10717 * doc/extend.texi (attribute const): Fix a typo.
10718 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10719 issuing -Wsuggest-attribute for void functions.
10721 2018-01-03 Martin Sebor <msebor@redhat.com>
10723 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10724 offset_int::from instead of wide_int::to_shwi.
10725 (maybe_diag_overlap): Remove assertion.
10726 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10727 * gimple-ssa-sprintf.c (format_directive): Same.
10728 (parse_directive): Same.
10729 (sprintf_dom_walker::compute_format_length): Same.
10730 (try_substitute_return_value): Same.
10732 2018-01-03 Jeff Law <law@redhat.com>
10734 PR middle-end/83654
10735 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10736 non-constant residual for zero at runtime and avoid probing in
10737 that case. Reorganize code for trailing problem to mirror handling
10740 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10742 PR tree-optimization/83501
10743 * tree-ssa-strlen.c (get_string_cst): New.
10744 (handle_char_store): Call get_string_cst.
10746 2018-01-03 Martin Liska <mliska@suse.cz>
10748 PR tree-optimization/83593
10749 * tree-ssa-strlen.c: Include tree-cfg.h.
10750 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10751 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10752 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10754 (strlen_dom_walker::before_dom_children): Call
10755 gimple_purge_dead_eh_edges. Dump tranformation with details
10757 (strlen_dom_walker::before_dom_children): Update call by adding
10758 new argument cleanup_eh.
10759 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10761 2018-01-03 Martin Liska <mliska@suse.cz>
10764 * cif-code.def (VARIADIC_THUNK): New enum value.
10765 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10768 2018-01-03 Jan Beulich <jbeulich@suse.com>
10770 * sse.md (mov<mode>_internal): Tighten condition for when to use
10771 vmovdqu<ssescalarsize> for TI and OI modes.
10773 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10775 Update copyright years.
10777 2018-01-03 Martin Liska <mliska@suse.cz>
10780 * ipa-visibility.c (function_and_variable_visibility): Skip
10781 functions with noipa attribure.
10783 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10785 * gcc.c (process_command): Update copyright notice dates.
10786 * gcov-dump.c (print_version): Ditto.
10787 * gcov.c (print_version): Ditto.
10788 * gcov-tool.c (print_version): Ditto.
10789 * gengtype.c (create_file): Ditto.
10790 * doc/cpp.texi: Bump @copying's copyright year.
10791 * doc/cppinternals.texi: Ditto.
10792 * doc/gcc.texi: Ditto.
10793 * doc/gccint.texi: Ditto.
10794 * doc/gcov.texi: Ditto.
10795 * doc/install.texi: Ditto.
10796 * doc/invoke.texi: Ditto.
10798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10800 * vector-builder.h (vector_builder::m_full_nelts): Change from
10801 unsigned int to poly_uint64.
10802 (vector_builder::full_nelts): Update prototype accordingly.
10803 (vector_builder::new_vector): Likewise.
10804 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10805 (vector_builder::operator ==): Likewise.
10806 (vector_builder::finalize): Likewise.
10807 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10808 Take the number of elements as a poly_uint64 rather than an
10810 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10811 from unsigned int to poly_uint64.
10812 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10813 (vec_perm_indices::new_vector): Likewise.
10814 (vec_perm_indices::length): Likewise.
10815 (vec_perm_indices::nelts_per_input): Likewise.
10816 (vec_perm_indices::input_nelts): Likewise.
10817 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10818 number of elements per input as a poly_uint64 rather than an
10819 unsigned int. Use the original encoding for variable-length
10820 vectors, rather than clamping each individual element.
10821 For the second and subsequent elements in each pattern,
10822 clamp the step and base before clamping their sum.
10823 (vec_perm_indices::series_p): Handle polynomial element counts.
10824 (vec_perm_indices::all_in_range_p): Likewise.
10825 (vec_perm_indices_to_tree): Likewise.
10826 (vec_perm_indices_to_rtx): Likewise.
10827 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10828 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10829 (tree_vector_builder::new_binary_operation): Handle polynomial
10830 element counts. Return false if we need to know the number
10831 of elements at compile time.
10832 * fold-const.c (fold_vec_perm): Punt if the number of elements
10833 isn't known at compile time.
10835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10837 * vec-perm-indices.h (vec_perm_builder): Change element type
10838 from HOST_WIDE_INT to poly_int64.
10839 (vec_perm_indices::element_type): Update accordingly.
10840 (vec_perm_indices::clamp): Handle polynomial element_types.
10841 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10842 (vec_perm_indices::all_in_range_p): Likewise.
10843 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10845 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10846 polynomial vec_perm_indices element types.
10847 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10848 * fold-const.c (fold_vec_perm): Likewise.
10849 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10850 * tree-vect-generic.c (lower_vec_perm): Likewise.
10851 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10852 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10853 element type to HOST_WIDE_INT.
10855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10856 Alan Hayward <alan.hayward@arm.com>
10857 David Sherwood <david.sherwood@arm.com>
10859 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10860 rather than an int. Use plus_constant.
10861 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10862 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10864 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10865 Alan Hayward <alan.hayward@arm.com>
10866 David Sherwood <david.sherwood@arm.com>
10868 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10869 a HOST_WIDE_INT to a poly_int64.
10871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10872 Alan Hayward <alan.hayward@arm.com>
10873 David Sherwood <david.sherwood@arm.com>
10875 * calls.c (load_register_parameters): Cope with polynomial
10876 mode sizes. Require a constant size for BLKmode parameters
10877 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10878 forces a parameter to be padded at the lsb end in order to
10879 fill a complete number of words, require the parameter size
10880 to be ordered wrt UNITS_PER_WORD.
10882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10883 Alan Hayward <alan.hayward@arm.com>
10884 David Sherwood <david.sherwood@arm.com>
10886 * reload1.c (spill_stack_slot_width): Change element type
10887 from unsigned int to poly_uint64_pod.
10888 (alter_reg): Treat mode sizes as polynomial.
10890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10891 Alan Hayward <alan.hayward@arm.com>
10892 David Sherwood <david.sherwood@arm.com>
10894 * reload.c (complex_word_subreg_p): New function.
10895 (reload_inner_reg_of_subreg, push_reload): Use it.
10897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10898 Alan Hayward <alan.hayward@arm.com>
10899 David Sherwood <david.sherwood@arm.com>
10901 * lra-constraints.c (process_alt_operands): Reject matched
10902 operands whose sizes aren't ordered.
10903 (match_reload): Refer to this check here.
10905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10906 Alan Hayward <alan.hayward@arm.com>
10907 David Sherwood <david.sherwood@arm.com>
10909 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10910 that the mode size is in the set {1, 2, 4, 8, 16}.
10912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10913 Alan Hayward <alan.hayward@arm.com>
10914 David Sherwood <david.sherwood@arm.com>
10916 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10917 Use plus_constant instead of gen_rtx_PLUS.
10919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10920 Alan Hayward <alan.hayward@arm.com>
10921 David Sherwood <david.sherwood@arm.com>
10923 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10924 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10925 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10926 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10927 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10928 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10929 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10930 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10931 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10932 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10934 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10935 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10936 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10937 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10938 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10939 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10940 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10941 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10942 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10943 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10945 * expr.c (emit_move_resolve_push): Treat the input and result
10946 of PUSH_ROUNDING as a poly_int64.
10947 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10948 (emit_push_insn): Likewise.
10949 * lra-eliminations.c (mark_not_eliminable): Likewise.
10950 * recog.c (push_operand): Likewise.
10951 * reload1.c (elimination_effects): Likewise.
10952 * rtlanal.c (nonzero_bits1): Likewise.
10953 * calls.c (store_one_arg): Likewise. Require the padding to be
10954 known at compile time.
10956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10957 Alan Hayward <alan.hayward@arm.com>
10958 David Sherwood <david.sherwood@arm.com>
10960 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10961 Use plus_constant instead of gen_rtx_PLUS.
10963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10964 Alan Hayward <alan.hayward@arm.com>
10965 David Sherwood <david.sherwood@arm.com>
10967 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10968 rather than an int.
10970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10971 Alan Hayward <alan.hayward@arm.com>
10972 David Sherwood <david.sherwood@arm.com>
10974 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10975 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10976 via stack temporaries. Treat the mode size as polynomial too.
10978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10979 Alan Hayward <alan.hayward@arm.com>
10980 David Sherwood <david.sherwood@arm.com>
10982 * expr.c (expand_expr_real_2): When handling conversions involving
10983 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10984 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10985 as a poly_uint64 too.
10987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10988 Alan Hayward <alan.hayward@arm.com>
10989 David Sherwood <david.sherwood@arm.com>
10991 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10994 Alan Hayward <alan.hayward@arm.com>
10995 David Sherwood <david.sherwood@arm.com>
10997 * combine.c (can_change_dest_mode): Handle polynomial
10998 REGMODE_NATURAL_SIZE.
10999 * expmed.c (store_bit_field_1): Likewise.
11000 * expr.c (store_constructor): Likewise.
11001 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11002 and polynomial REGMODE_NATURAL_SIZE.
11003 (gen_lowpart_common): Likewise.
11004 * reginfo.c (record_subregs_of_mode): Likewise.
11005 * rtlanal.c (read_modify_subreg_p): Likewise.
11007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11008 Alan Hayward <alan.hayward@arm.com>
11009 David Sherwood <david.sherwood@arm.com>
11011 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11012 numbers of elements.
11014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11015 Alan Hayward <alan.hayward@arm.com>
11016 David Sherwood <david.sherwood@arm.com>
11018 * match.pd: Cope with polynomial numbers of vector elements.
11020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11021 Alan Hayward <alan.hayward@arm.com>
11022 David Sherwood <david.sherwood@arm.com>
11024 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11025 in a POINTER_PLUS_EXPR.
11027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11028 Alan Hayward <alan.hayward@arm.com>
11029 David Sherwood <david.sherwood@arm.com>
11031 * omp-simd-clone.c (simd_clone_subparts): New function.
11032 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11033 (ipa_simd_modify_function_body): Likewise.
11035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11036 Alan Hayward <alan.hayward@arm.com>
11037 David Sherwood <david.sherwood@arm.com>
11039 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11040 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11041 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11042 (expand_vector_condition, vector_element): Likewise.
11043 (subparts_gt): New function.
11044 (get_compute_type): Use subparts_gt.
11045 (count_type_subparts): Delete.
11046 (expand_vector_operations_1): Use subparts_gt instead of
11047 count_type_subparts.
11049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11050 Alan Hayward <alan.hayward@arm.com>
11051 David Sherwood <david.sherwood@arm.com>
11053 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11054 (vect_compile_time_alias): ...this new function. Do the calculation
11055 on poly_ints rather than trees.
11056 (vect_prune_runtime_alias_test_list): Update call accordingly.
11058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11059 Alan Hayward <alan.hayward@arm.com>
11060 David Sherwood <david.sherwood@arm.com>
11062 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11064 (vect_schedule_slp_instance): Likewise.
11066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11067 Alan Hayward <alan.hayward@arm.com>
11068 David Sherwood <david.sherwood@arm.com>
11070 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11071 constant and extern definitions for variable-length vectors.
11072 (vect_get_constant_vectors): Note that the number of units
11073 is known to be constant.
11075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11076 Alan Hayward <alan.hayward@arm.com>
11077 David Sherwood <david.sherwood@arm.com>
11079 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11080 of units as polynomial. Choose between WIDE and NARROW based
11083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11084 Alan Hayward <alan.hayward@arm.com>
11085 David Sherwood <david.sherwood@arm.com>
11087 * tree-vect-stmts.c (simd_clone_subparts): New function.
11088 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11091 Alan Hayward <alan.hayward@arm.com>
11092 David Sherwood <david.sherwood@arm.com>
11094 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11095 vectors as polynomial. Use build_index_vector for
11096 IFN_GOMP_SIMD_LANE.
11098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11099 Alan Hayward <alan.hayward@arm.com>
11100 David Sherwood <david.sherwood@arm.com>
11102 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11103 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11104 for variable-length vectors.
11105 (vectorizable_mask_load_store): Treat the number of units as
11106 polynomial, asserting that it is constant if the condition has
11107 already been enforced.
11108 (vectorizable_store, vectorizable_load): Likewise.
11110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11111 Alan Hayward <alan.hayward@arm.com>
11112 David Sherwood <david.sherwood@arm.com>
11114 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11115 of units as polynomial. Punt if we can't tell at compile time
11116 which vector contains the final result.
11118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11119 Alan Hayward <alan.hayward@arm.com>
11120 David Sherwood <david.sherwood@arm.com>
11122 * tree-vect-loop.c (vectorizable_induction): Treat the number
11123 of units as polynomial. Punt on SLP inductions. Use an integer
11124 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11125 cast of such a series for variable-length floating-point
11128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11129 Alan Hayward <alan.hayward@arm.com>
11130 David Sherwood <david.sherwood@arm.com>
11132 * tree.h (build_index_vector): Declare.
11133 * tree.c (build_index_vector): New function.
11134 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11135 of units as polynomial, forcibly converting it to a constant if
11136 vectorizable_reduction has already enforced the condition.
11137 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11138 to create a {1,2,3,...} vector.
11139 (vectorizable_reduction): Treat the number of units as polynomial.
11140 Choose vectype_in based on the largest scalar element size rather
11141 than the smallest number of units. Enforce the restrictions
11144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11145 Alan Hayward <alan.hayward@arm.com>
11146 David Sherwood <david.sherwood@arm.com>
11148 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11149 number of units as polynomial.
11151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11152 Alan Hayward <alan.hayward@arm.com>
11153 David Sherwood <david.sherwood@arm.com>
11155 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11156 * target.def (autovectorize_vector_sizes): Return the vector sizes
11157 by pointer, using vector_sizes rather than a bitmask.
11158 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11159 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11160 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11162 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11163 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11164 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11165 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11166 * omp-general.c (omp_max_vf): Likewise.
11167 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11168 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11169 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11170 * tree-vect-slp.c (vect_slp_bb): Likewise.
11171 * doc/tm.texi: Regenerate.
11172 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11174 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11175 the vector size as a poly_uint64 rather than an unsigned int.
11176 (current_vector_size): Change from an unsigned int to a poly_uint64.
11177 (get_vectype_for_scalar_type): Update accordingly.
11178 * tree.h (build_truth_vector_type): Take the size and number of
11179 units as a poly_uint64 rather than an unsigned int.
11180 (build_vector_type): Add a temporary overload that takes
11181 the number of units as a poly_uint64 rather than an unsigned int.
11182 * tree.c (make_vector_type): Likewise.
11183 (build_truth_vector_type): Take the number of units as a poly_uint64
11184 rather than an unsigned int.
11186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11187 Alan Hayward <alan.hayward@arm.com>
11188 David Sherwood <david.sherwood@arm.com>
11190 * target.def (get_mask_mode): Take the number of units and length
11191 as poly_uint64s rather than unsigned ints.
11192 * targhooks.h (default_get_mask_mode): Update accordingly.
11193 * targhooks.c (default_get_mask_mode): Likewise.
11194 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11195 * doc/tm.texi: Regenerate.
11197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11198 Alan Hayward <alan.hayward@arm.com>
11199 David Sherwood <david.sherwood@arm.com>
11201 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11202 * omp-general.c (omp_max_vf): Likewise.
11203 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11204 (expand_omp_simd): Handle polynomial safelen.
11205 * omp-low.c (omplow_simd_context): Add a default constructor.
11206 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11207 (lower_rec_simd_input_clauses): Update accordingly.
11208 (lower_rec_input_clauses): Likewise.
11210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11211 Alan Hayward <alan.hayward@arm.com>
11212 David Sherwood <david.sherwood@arm.com>
11214 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11215 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11216 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11217 (vect_analyze_slp_cost): Likewise.
11218 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11219 (vect_model_load_cost): Likewise.
11221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11222 Alan Hayward <alan.hayward@arm.com>
11223 David Sherwood <david.sherwood@arm.com>
11225 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11226 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11227 from an unsigned int * to a poly_uint64_pod *.
11228 (calculate_unrolling_factor): New function.
11229 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11232 Alan Hayward <alan.hayward@arm.com>
11233 David Sherwood <david.sherwood@arm.com>
11235 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11236 from an unsigned int to a poly_uint64.
11237 (_loop_vec_info::slp_unrolling_factor): Likewise.
11238 (_loop_vec_info::vectorization_factor): Change from an int
11240 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11241 (vect_get_num_vectors): New function.
11242 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11243 (vect_get_num_copies): Use vect_get_num_vectors.
11244 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11245 to an unsigned int *.
11246 (vect_analyze_data_refs): Change min_vf from an int * to a
11248 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11249 than an unsigned HOST_WIDE_INT.
11250 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11251 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11252 to an unsigned int *.
11253 (vect_analyze_data_ref_dependences): Likewise.
11254 (vect_compute_data_ref_alignment): Handle polynomial vf.
11255 (vect_enhance_data_refs_alignment): Likewise.
11256 (vect_prune_runtime_alias_test_list): Likewise.
11257 (vect_shift_permute_load_chain): Likewise.
11258 (vect_supportable_dr_alignment): Likewise.
11259 (dependence_distance_ge_vf): Take the vectorization factor as a
11260 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11261 (vect_analyze_data_refs): Change min_vf from an int * to a
11263 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11264 vfm1 as a poly_uint64 rather than an int. Make the same change
11265 for the returned bound_scalar.
11266 (vect_gen_vector_loop_niters): Handle polynomial vf.
11267 (vect_do_peeling): Likewise. Update call to
11268 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11269 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11271 * tree-vect-loop.c (vect_determine_vectorization_factor)
11272 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11273 (vect_get_known_peeling_cost): Likewise.
11274 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11275 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11276 (vect_transform_loop): Likewise. Use the lowest possible VF when
11277 updating the upper bounds of the loop.
11278 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11279 rather than an int.
11280 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11281 polynomial unroll factors.
11282 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11283 (vect_make_slp_decision): Likewise.
11284 (vect_supported_load_permutation_p): Likewise, and polynomial
11286 (vect_analyze_slp_cost): Handle polynomial vf.
11287 (vect_slp_analyze_node_operations): Likewise.
11288 (vect_slp_analyze_bb_1): Likewise.
11289 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11290 than an unsigned HOST_WIDE_INT.
11291 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11292 (vectorizable_load): Handle polynomial vf.
11293 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11295 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11298 Alan Hayward <alan.hayward@arm.com>
11299 David Sherwood <david.sherwood@arm.com>
11301 * match.pd: Handle bit operations involving three constants
11302 and try to fold one pair.
11304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11306 * tree-vect-loop-manip.c: Include gimple-fold.h.
11307 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11308 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11309 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11310 Add a path that uses a step of VF instead of 1, but disable it
11312 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11313 and niters_no_overflow parameters. Update calls to
11314 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11315 Create a new SSA name if the latter choses to use a ste other
11316 than zero, and return it via niters_vector_mult_vf_var.
11317 * tree-vect-loop.c (vect_transform_loop): Update calls to
11318 vect_do_peeling, vect_gen_vector_loop_niters and
11319 slpeel_make_loop_iterate_ntimes.
11320 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11321 (vect_gen_vector_loop_niters): Update declarations after above changes.
11323 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11325 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11326 128-bit round to integer instructions.
11327 (ceil<mode>2): Likewise.
11328 (btrunc<mode>2): Likewise.
11329 (round<mode>2): Likewise.
11331 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11333 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11334 unaligned VSX load/store on P8/P9.
11335 (expand_block_clear): Allow the use of unaligned VSX
11336 load/store on P8/P9.
11338 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11340 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11342 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11343 swap associated with both a load and a store.
11345 2018-01-02 Andrew Waterman <andrew@sifive.com>
11347 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11348 * config/riscv/riscv.md (clear_cache): Use it.
11350 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11352 * web.c: Remove out-of-date comment.
11354 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11356 * expr.c (fixup_args_size_notes): Check that any existing
11357 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11358 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11359 (emit_single_push_insn): ...here.
11361 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11363 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11364 (const_vector_encoded_nelts): New function.
11365 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11366 (const_vector_int_elt, const_vector_elt): Declare.
11367 * emit-rtl.c (const_vector_int_elt_1): New function.
11368 (const_vector_elt): Likewise.
11369 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11370 of CONST_VECTOR_ELT.
11372 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11374 * expr.c: Include rtx-vector-builder.h.
11375 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11376 directly on the tree encoding.
11377 (const_vector_from_tree): Likewise.
11378 * optabs.c: Include rtx-vector-builder.h.
11379 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11380 sequence of "u" values.
11381 * vec-perm-indices.c: Include rtx-vector-builder.h.
11382 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11383 directly on the vec_perm_indices encoding.
11385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11387 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11388 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11389 * rtx-vector-builder.h: New file.
11390 * rtx-vector-builder.c: Likewise.
11391 * rtl.h (rtx_def::u2): Add a const_vector field.
11392 (CONST_VECTOR_NPATTERNS): New macro.
11393 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11394 (CONST_VECTOR_DUPLICATE_P): Likewise.
11395 (CONST_VECTOR_STEPPED_P): Likewise.
11396 (CONST_VECTOR_ENCODED_ELT): Likewise.
11397 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11398 (unwrap_const_vec_duplicate): Likewise.
11399 (const_vec_series_p): Check for a non-duplicated vector encoding.
11400 Say that the function only returns true for integer vectors.
11401 * emit-rtl.c: Include rtx-vector-builder.h.
11402 (gen_const_vec_duplicate_1): Delete.
11403 (gen_const_vector): Call gen_const_vec_duplicate instead of
11404 gen_const_vec_duplicate_1.
11405 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11406 (gen_const_vec_duplicate): Use rtx_vector_builder.
11407 (gen_const_vec_series): Likewise.
11408 (gen_rtx_CONST_VECTOR): Likewise.
11409 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11410 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11411 Build a new vector rather than modifying a CONST_VECTOR in-place.
11412 (handle_special_swappables): Update call accordingly.
11413 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11414 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11415 Build a new vector rather than modifying a CONST_VECTOR in-place.
11416 (handle_special_swappables): Update call accordingly.
11418 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11420 * simplify-rtx.c (simplify_const_binary_operation): Use
11421 CONST_VECTOR_ELT instead of XVECEXP.
11423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11425 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11426 the selector elements to be different from the data elements
11427 if the selector is a VECTOR_CST.
11428 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11429 ssizetype for the selector.
11431 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11433 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11434 before testing each element individually.
11435 * tree-vect-generic.c (lower_vec_perm): Likewise.
11437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11439 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11440 * selftest-run-tests.c (selftest::run_tests): Call it.
11441 * vector-builder.h (vector_builder::operator ==): New function.
11442 (vector_builder::operator !=): Likewise.
11443 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11444 (vec_perm_indices::all_from_input_p): New function.
11445 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11446 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11447 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11448 instead of reading the VECTOR_CST directly. Detect whether both
11449 vector inputs are the same before constructing the vec_perm_indices,
11450 and update the number of inputs argument accordingly. Use the
11451 utility functions added above. Only construct sel2 if we need to.
11453 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11455 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11456 the broadcast of the low byte.
11457 (expand_mult_highpart): Use an explicit encoding for the permutes.
11458 * optabs-query.c (can_mult_highpart_p): Likewise.
11459 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11460 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11461 (vectorizable_bswap): Likewise.
11462 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11463 explicit encoding for the power-of-2 permutes.
11464 (vect_permute_store_chain): Likewise.
11465 (vect_grouped_load_supported): Likewise.
11466 (vect_permute_load_chain): Likewise.
11468 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11470 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11471 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11472 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11473 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11474 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11475 (vect_gen_perm_mask_any): Likewise.
11477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11479 * int-vector-builder.h: New file.
11480 * vec-perm-indices.h: Include int-vector-builder.h.
11481 (vec_perm_indices): Redefine as an int_vector_builder.
11482 (auto_vec_perm_indices): Delete.
11483 (vec_perm_builder): Redefine as a stand-alone class.
11484 (vec_perm_indices::vec_perm_indices): New function.
11485 (vec_perm_indices::clamp): Likewise.
11486 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11487 (vec_perm_indices::new_vector): New function.
11488 (vec_perm_indices::new_expanded_vector): Update for new
11489 vec_perm_indices class.
11490 (vec_perm_indices::rotate_inputs): New function.
11491 (vec_perm_indices::all_in_range_p): Operate directly on the
11492 encoded form, without computing elided elements.
11493 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11494 encoding. Update for new vec_perm_indices class.
11495 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11496 the given vec_perm_builder.
11497 (expand_vec_perm_var): Update vec_perm_builder constructor.
11498 (expand_mult_highpart): Use vec_perm_builder instead of
11499 auto_vec_perm_indices.
11500 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11501 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11502 or double series encoding as appropriate.
11503 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11504 vec_perm_indices instead of auto_vec_perm_indices.
11505 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11506 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11507 (vect_permute_store_chain): Likewise.
11508 (vect_grouped_load_supported): Likewise.
11509 (vect_permute_load_chain): Likewise.
11510 (vect_shift_permute_load_chain): Likewise.
11511 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11512 (vect_transform_slp_perm_load): Likewise.
11513 (vect_schedule_slp_instance): Likewise.
11514 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11515 (vectorizable_mask_load_store): Likewise.
11516 (vectorizable_bswap): Likewise.
11517 (vectorizable_store): Likewise.
11518 (vectorizable_load): Likewise.
11519 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11520 vec_perm_indices instead of auto_vec_perm_indices. Use
11521 tree_to_vec_perm_builder to read the vector from a tree.
11522 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11523 vec_perm_builder instead of a vec_perm_indices.
11524 (have_whole_vector_shift): Use vec_perm_builder and
11525 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11526 truncation to calc_vec_perm_mask_for_shift.
11527 (vect_create_epilog_for_reduction): Likewise.
11528 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11529 from auto_vec_perm_indices to vec_perm_indices.
11530 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11531 instead of changing individual elements.
11532 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11533 the vector in d.perm.
11534 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11535 from auto_vec_perm_indices to vec_perm_indices.
11536 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11537 instead of changing individual elements.
11538 (arm_vectorize_vec_perm_const): Use new_vector to install
11539 the vector in d.perm.
11540 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11541 Update vec_perm_builder constructor.
11542 (rs6000_expand_interleave): Likewise.
11543 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11544 (rs6000_expand_interleave): Likewise.
11546 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11548 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11549 to qimode could truncate the indices.
11550 * optabs.c (expand_vec_perm_var): Likewise.
11552 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11554 * Makefile.in (OBJS): Add vec-perm-indices.o.
11555 * vec-perm-indices.h: New file.
11556 * vec-perm-indices.c: Likewise.
11557 * target.h (vec_perm_indices): Replace with a forward class
11559 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11560 * optabs.h: Include vec-perm-indices.h.
11561 (expand_vec_perm): Delete.
11562 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11563 (expand_vec_perm_const): Declare.
11564 * target.def (vec_perm_const_ok): Replace with...
11565 (vec_perm_const): ...this new hook.
11566 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11567 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11568 * doc/tm.texi: Regenerate.
11569 * optabs.def (vec_perm_const): Delete.
11570 * doc/md.texi (vec_perm_const): Likewise.
11571 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11572 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11573 expand_vec_perm for constant permutation vectors. Assert that
11574 the mode of variable permutation vectors is the integer equivalent
11575 of the mode that is being permuted.
11576 * optabs-query.h (selector_fits_mode_p): Declare.
11577 * optabs-query.c: Include vec-perm-indices.h.
11578 (selector_fits_mode_p): New function.
11579 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11580 is defined, instead of checking whether the vec_perm_const_optab
11581 exists. Use targetm.vectorize.vec_perm_const instead of
11582 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11583 fit in the vector mode before using a variable permute.
11584 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11585 vec_perm_indices instead of an rtx.
11586 (expand_vec_perm): Replace with...
11587 (expand_vec_perm_const): ...this new function. Take the selector
11588 as a vec_perm_indices rather than an rtx. Also take the mode of
11589 the selector. Update call to shift_amt_for_vec_perm_mask.
11590 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11591 Use vec_perm_indices::new_expanded_vector to expand the original
11592 selector into bytes. Check whether the indices fit in the vector
11593 mode before using a variable permute.
11594 (expand_vec_perm_var): Make global.
11595 (expand_mult_highpart): Use expand_vec_perm_const.
11596 * fold-const.c: Includes vec-perm-indices.h.
11597 * tree-ssa-forwprop.c: Likewise.
11598 * tree-vect-data-refs.c: Likewise.
11599 * tree-vect-generic.c: Likewise.
11600 * tree-vect-loop.c: Likewise.
11601 * tree-vect-slp.c: Likewise.
11602 * tree-vect-stmts.c: Likewise.
11603 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11605 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11606 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11607 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11608 (aarch64_vectorize_vec_perm_const): ...this new function.
11609 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11610 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11611 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11612 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11613 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11614 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11615 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11617 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11618 check for NEON modes.
11619 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11620 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11621 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11622 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11624 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11625 the old VEC_PERM_CONST conditions.
11626 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11627 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11628 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11629 (ia64_vectorize_vec_perm_const_ok): Merge into...
11630 (ia64_vectorize_vec_perm_const): ...this new function.
11631 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11632 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11633 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11634 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11635 * config/mips/mips.c (mips_expand_vec_perm_const)
11636 (mips_vectorize_vec_perm_const_ok): Merge into...
11637 (mips_vectorize_vec_perm_const): ...this new function.
11638 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11639 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11640 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11641 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11642 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11643 (rs6000_expand_vec_perm_const): Delete.
11644 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11646 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11647 (altivec_expand_vec_perm_const_le): Take each operand individually.
11648 Operate on constant selectors rather than rtxes.
11649 (altivec_expand_vec_perm_const): Likewise. Update call to
11650 altivec_expand_vec_perm_const_le.
11651 (rs6000_expand_vec_perm_const): Delete.
11652 (rs6000_vectorize_vec_perm_const_ok): Delete.
11653 (rs6000_vectorize_vec_perm_const): New function.
11654 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11655 an element count and rtx array.
11656 (rs6000_expand_extract_even): Update call accordingly.
11657 (rs6000_expand_interleave): Likewise.
11658 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11659 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11660 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11661 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11662 (rs6000_expand_vec_perm_const): Delete.
11663 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11664 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11665 (altivec_expand_vec_perm_const_le): Take each operand individually.
11666 Operate on constant selectors rather than rtxes.
11667 (altivec_expand_vec_perm_const): Likewise. Update call to
11668 altivec_expand_vec_perm_const_le.
11669 (rs6000_expand_vec_perm_const): Delete.
11670 (rs6000_vectorize_vec_perm_const_ok): Delete.
11671 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11672 reference to the SPE evmerge intructions.
11673 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11674 an element count and rtx array.
11675 (rs6000_expand_extract_even): Update call accordingly.
11676 (rs6000_expand_interleave): Likewise.
11677 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11678 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11680 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11682 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11684 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11685 vector mode and that that mode matches the mode of the data
11687 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11688 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11689 directly using expand_vec_perm_1 when forcing selectors into
11691 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11695 * optabs-query.h (can_vec_perm_p): Delete.
11696 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11697 * optabs-query.c (can_vec_perm_p): Split into...
11698 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11699 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11700 particular selector is valid.
11701 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11702 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11703 (vect_grouped_load_supported): Likewise.
11704 (vect_shift_permute_load_chain): Likewise.
11705 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11706 (vect_transform_slp_perm_load): Likewise.
11707 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11708 (vectorizable_bswap): Likewise.
11709 (vect_gen_perm_mask_checked): Likewise.
11710 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11711 implementations of variable permutation vectors into account
11712 when deciding which selector to use.
11713 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11714 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11715 with a false third argument.
11716 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11717 to test whether the constant selector is valid and can_vec_perm_var_p
11718 to test whether a variable selector is valid.
11720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11722 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11723 * optabs-query.c (can_vec_perm_p): Likewise.
11724 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11725 instead of vec_perm_indices.
11726 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11727 (vect_gen_perm_mask_checked): Likewise,
11728 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11729 (vect_gen_perm_mask_checked): Likewise,
11731 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11733 * optabs-query.h (qimode_for_vec_perm): Declare.
11734 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11735 (qimode_for_vec_perm): ...this new function.
11736 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11738 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11740 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11741 does not have a conditional at the top.
11743 2018-01-02 Richard Biener <rguenther@suse.de>
11745 * ipa-inline.c (big_speedup_p): Fix expression.
11747 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11750 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11753 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11757 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11758 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11759 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11760 cond_taken_branch_cost 3->4.
11762 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11764 PR tree-optimization/83581
11765 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11766 TODO_cleanup_cfg if any changes have been made.
11768 PR middle-end/83608
11769 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11770 convert_modes if target mode has the right side, but different mode
11773 PR middle-end/83609
11774 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11775 last argument when extracting from CONCAT. If either from_real or
11776 from_imag is NULL, use expansion through memory. If result is not
11777 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11778 the parts directly to inner mode, if even that fails, use expansion
11781 PR middle-end/83623
11782 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11783 check for bswap in mode rather than HImode and use that in expand_unop
11786 Copyright (C) 2018 Free Software Foundation, Inc.
11788 Copying and distribution of this file, with or without modification,
11789 are permitted in any medium without royalty provided the copyright
11790 notice and this notice are preserved.